Bringing AI to maritime
79 min • Published Jan 15, 2024
Rate this video
00:00:00Introduction and Background
The show begins with an introduction of the episode's focus on maritime innovation and artificial intelligence (AI). The host introduces Tony and Ben from Cedo AI, a company that aims to revolutionize the maritime industry by optimizing vessel performance using AI.00:13:17The Story Behind Cedo AI
The conversation explores the story behind Cedo AI, the challenges they've faced, their successes, and their vision for the future of AI and maritime. Tony explains the formation and concept behind his company, Sito AI, and how it aims to solve problems in the maritime industry using existing data streams produced by vessels.00:26:33Importance of Monitoring and Maintaining Maritime Vessels
The discussion shifts to the importance of monitoring and maintaining maritime vessels, specifically their engines. The speaker discusses how traditionally, the majority of the analysis was done by the engineers on board, but with the advancement of technology, remote data analysis has become more prevalent.00:39:49Data Collection and Analysis in Maritime Operations
The speaker discusses the process of data collection and analysis in the context of maritime operations. They mention the challenges of avoiding duplicate data and the need for effective data compression techniques. The speaker also discusses the importance of predicting potential catastrophic failures before they happen.00:53:05Use of AI and Data Analytics in Maritime Operations
The conversation delves into the technical aspects of using AI and data analytics in maritime operations. Ben explains how they use AI to track and analyze data from vessels, extending the prediction window and allowing crew members to intervene at the most convenient and optimal time.01:06:21Challenges and Solutions of a Maritime Startup
The conversation concludes with a discussion on the challenges and solutions of a maritime startup that uses MongoDB. The startup collects data from vessels and transmits it to the cloud for analysis. They discuss the measures they've implemented to overcome challenges and ensure the security and privacy of the data.The primary focus of the video is the application of artificial intelligence in the maritime industry, specifically through the work of Cedo AI, and the role of MongoDB in managing and analyzing large amounts of data.
🔑 Key Points
- Cedo AI aims to revolutionize the maritime industry by optimizing vessel performance using AI.
- The company uses AI to leverage data, using supervised learning to understand what the vessel is doing and running anomaly detection to identify any unusual patterns.
- MongoDB is used for their data due to its good handling of time series data and sensible approach to compression.
- The company is set to launch a new product to help with EU ETS (European Union's emission trading scheme) compliance.
- The company aims to have their technology on 70-80 vessels by the end of the year.
- The company recently raised a one and a half million pound seed round led by Howen Ventures.
🔗 Related Links
Full Video Transcript
[Music] [Music] [Music] [Music] [Music] [Music] well it's Thursday so as ever welcome to another episode of mongod TV Cloud connect this is the first show of 2024 so welcome back to all our regular viewers and to all our new viewers you're very very welcome Maritime Innovation is at the Forefront of our discussion today as we Deep dive into the world of artificial intelligence and its transformative impact on the Seas but but before we start let's do a little bit of housekeeping for those who don't know me I'm Shane mallister I'm a lead and the developer relations team here at mongodb and I host this show every Thursday on YouTube and on LinkedIn so it's great to have you back if you've joined us before and if you're new you're very welcome you can go to LinkedIn or YouTube to see all of the past shows in our Cloud connect series and indeed you'll see some of our forthcoming episodes as well too so while we're waiting to get underway with the show do let us know in the comments on LinkedIn and YouTube who you are and where you're tuning in from we'd love to hear from you and once we get underway if you have any questions for our guests or for me drop them in the chat and we'll try to answer them live or if we don't get an opportunity to do so live we will cover them at the end this is being recorded So and it will be available at mong's YouTube Channel Once finished so don't worry if you can't stay for all of it um it the recording will be there for you to jump back into and if you're on YouTube don't forget to like And subscribe if you haven't already done so and why haven't you quite honestly and if you are on LinkedIn please follow mongod B for all the latest news and posts from us so with all the housekeeping out of the way let's get on with the show and I'm super excited we've delayed this show a couple of times before so I'm thrilled that we're finally getting to broadcast this so today I'm joined by Tony hildrew the CEO and founder and Ben Harrison the CTO of cedo AI based in Newcastle cedo AI is on a mission to revolutionize the maritime industry by harnessing the power of AI to optimize vessel performance I suppose the vast oceans that connect our world have been a playground for Innovation and cedo is no different than that they're navigating these Waters pardon the puns with cuttingedge solutions that will reshape how ships operate and perform so we're going to uncover the story behind C aai today the challenges they' faced founding the company the successes that they've achieved so far and the exciting future that they Envision for the intersection of AI and Maritime we're going to Deep dive into the intricacies of AI technology and how that applies to the high seas and explore the data driven insights that Propel the maritime industry into a new area from cargo vessels to cruise liners the impact of what they're doing is far-reaching promising increased efficiency reduced environmental impact and above all safer voyages so without further Ado let's join Tony and Ben on the stream and uh you're very both very welcome to mongodb Cloud connect how are you very good thank you thanks very much for having us I know uh as you mentioned it's been a long time coming but uh very happy to be here yeah I'm I'm I'm thrilled we got this one off the ground I think we were talking first a couple of months ago and certain things intervened and sicknesses Etc as well too so it's great that you you hung out I suppose before we get into cedo as a company Tony and Ben perhaps maybe tell us what your backgrounds are how you got here and what your day-to-day roles are in cedo so Tony if I could start with you please yeah um I mean myself personally I live and breathe Maritime uh so fresh out of high school I pursued a career in the maritime industry um starting off as a marine engineer so worked at se um on board a number of different types of vessels bulk carriers uh passenger ships and then also a short period of time in the uh super yacht sector so the bit more glamorous side of things um compared to the conventional commcial and iron all um definitely so then I uh I left once I left that industry I moved back ashore and worked as a uh project manager project engineer in a shipyard doing um new build and uh refit projects on a a bunch of containerships before taking up the vessel manager role where uh I managed to Fleet a 50 15 vessels and it was there that I really identified the problem but I'll put the breaks on for now and and let Ben introduce himself a little bit more yeah so Ben Harrison CTO here uh I suppose live and breathe startups in the same way that Tony lives in breathes Maritime I started my first startup during University um went on to uh work there for five years um grew it very quickly organically um before actually that all ended up in tears five years later uh and uh spent the the next number of years working in um effectively Contracting and Consulting roles paying off the the the personal loans and debt that had uh accured from from the previous position um background in yeah operationally intensive and and software uh intensive startup um and smmes and uh joined Tony here back uh sort of start mid last year um to to start building out the the the tech stack from the MVP that that Tony and uh the the previous uh Dev had had built out excellent excellent I I presume from both of those introductions we could probably do another episode on Tony's life in the luxury yacht sector and Ben's Horror Story in startup land right so but let's park those for the moment so Sito how did it come about um I suppose I'll start with you Tony how did it come about obviously with your introductions there you were ingrained in the industry you saw a problem you saw an issue you saw a challenge to be solved so how did the formation come about and also tell us the interesting story behind the name in the first place yeah so um I mean the way that the company was initially formed was uh it was during my time as a vessel manager and I was facing issues around um insurance policy renewals of all things uh speaking with a number of Brokers and I had multiple requests for repeats of information and it was just getting very timec consuming very tedious and I thought to myself okay there's got to be a quicker way to do that can we speed up that submissions process can we obtain Insurance quicker cheaper fairer all the rest of it and then I started to think some of the problems that I faced as an engineer probably played heavily into the pricing and understanding of risk did some research and fortunately for us there was no other solution around that was tapping into existing data streams which um vessels produce uh and using it for insurance purposes so if you're not familiar with a ship which most people aren't not a lot of people get especially into the Machinery spaces of these things they've got an abundance of sensors and they're constantly monitoring things like pressures temperatures loads voltages you name it it's probably got a sensor attached to it everything's relayed back to an alarm and monitoring system and um the idea essentially was as an engineer you take all of that information you digest it and you diagnose the problem so why don't we do that remotely and that would essentially speed up and solve some of the problems I was facing in my day job as vessel manager but it also solves the problems of the engineer and I cast my mind back to the time where you know you've been at work from 8 until 5: you just settle down for the night and then all of a sudden you get a um an alarm and you have to go back down to the engine room and on behold you discover that a cylinder liner needs to be pulled so you're there and you know that that's about a 12-hour job if you do it fast if you do it efficient it's it's very tiring it's hard work um so if we can make the live to see far is easier it also makes it safer from a fatigue perspective as well and and that's where the ideas really started to come together I then looked into the commerci is um you know if that vessel is out of service so you take a container vessel as an example costs typically $30,000 per day to Charter Plus another $65,000 in operational costs that's significant amount of money to lose if a vessel's out of operation for just a single day if we can identify these mechanical issues that are going to cause vessel downtime ahead of time action can be taken um at the most convenient time to prevent that cost being incurred and then I supposed to go on to your second question about where the name came from we were actually called uh signal intelligence prior to to being called cedo and um you know as as small companies you kind of try to navigate a lot of different Landscapes and uh yeah signal intelligence was unfortunately taken but I think we've landed on a better name and the way that the name came up was I was in the car with my partner and we were discussing you know name ideas and how we can how we can sort of find something that fits really well um and Sito was actually a Greek goddess and she was a Greek goddess of hidden dangers of the sea and all all of a sudden just clicked and it's was like okay that's it that's the one uh you know let's nail it to the wall and and from then on uh seeo AI That's so excellent that's the company and most importantly you could get the domain as well too which is always pretty tricky in many respects but super tricky for short names right yeah I mean we were very lucky with that I think uh the we we did land on a fantastic domain name as well it can't get any easier than z. a yeah that's brilliant I love how you painted the picture for us there because as you rightly say I mean we probably all maybe all and I'm looking at at the comments that we're getting where people joining from you know all sides of the globe and it is a truly Global industry and it's great to see people from the US and Mexico and Barcelona and Canada and even Dublin just up the road from where I am as well too um and India of course as well um as you said not many of us might know the intricacies of the maritime industry we might have been a passenger on a ferry or a cruise ship or something like that you touched a little bit on the scale there of the cost downtime has a direct intrinsic cost and I think you know for me and when we talk to guests at the come on cloud connect that's always something that maybe from a developers perspective we're not exposed to so much we're you know downtime in in a digital space is you know companies talk about 99.9% uptime and you know there's always a failover etc but I can imagine in a maritime case you don't have but maybe you have other engines but it's going slower Etc what's the failover is it just it stops you have to down tools and everybody hands on deck to fix it yeah I mean that's essentially it you know if if you have a major failure with um say for example you take a large bul carrier it has one main engine which is a direct drive so you have a single shaft that comes out with a propeller attached to the end end of it you've got an issue with an exhaust gas valve or um a cylinder liner you know there's there's only certain amount of things you can do one of the elements in in really bad instances you could lift the fuel pump and stop using that unit obviously that results in lower power the other option that you have is you stop the engine you wait for to cool down and then you you know ventilate the spaces that you need access to and you conduct the maintenance on it and you know to put it into perspective some of these engines that they're almost the same size as a house they're two stories high um they produce an awful lot of power you know a small slow speed engine can produce 10,000 kilowatts of of power um so from that side of things it's it it's a Paramount importance that these these engines are well maintained um but also the the auxiliary systems that support those engines you know you've got pumps you've got uh heat exchanges um other cooling water systems which which need to be functioning correctly in order for the main power unit to actually Propel The Vessel through the water and the other thing that that just to tack on the end there um in terms of the the impact of having a problem you know maritime's at the center of global trade and so it's not just an isolated event we all saw a couple of years back what happened when the sewers Canal got blocked you know these these events do have significant Downstream effects Beyond just the maritime industry itself yeah and I think you know we we we feel that and even look it's ongoing at the moment there is issues there as well too with with the SE Canal Etc um you mentioned the sensors and the alarms and I presume some control panel board with flashing lights somewhere what was the landscape of of monitoring and checking for these potential faults before you founded the company how you know who who was in The Who were the main players how are things traditionally done was it just that sensors alarms flashing lights and and warnings yeah so um I mean when I first started out in Maritime back in 2011 there you know that that point in time you had your senses but the majority of the analysis was done by the engineers on board um you get daily reports which are sent back to the Shor side staff uh commercial and Technical teams but the majority of the diagnosis is done on board now as you you know as things have moved on over the last decade um you have seen people come in and and start to look data remotely um some of the solutions that have been deployed over the past few years of you know they capture data at a 6 minute to 10 minute interval which is not really as high frequency enough as what you would require when you you consider that diesel generators can be rotating at you know 1,800 RPM so see at a six minute interval you're not really going to Capt capture the information you need at a high enough Fidelity in order to actually act upon it now we've created a product that captures data at one one Hertz so every second um and and starts to lean more into that real time understanding of what's Happening um so in in one regard we've taken you know some solutions that have been present and and now we we've redeveloped those and refined what can be done um capturing data at a higher frequency you know making use of new technologies that are coming into play with uh Leo satellite constellations connecting vessels uh it it's completely changing the way that the industry will approach data um moving forward and and hopefully you know moving into the next 10 20 years as you look towards things like autonomous vessels and man vessels sure so I understand from my Layman's perspective of you know yes you're going to put these sensors all over the critical components Within inside the vessel what are the sensors actually measuring or looking out for before we get into how you Analyze That and into the AI piece of your stack and your technology what are they looking out for Ben so um the sensors generally are already installed so this is the the the big thing that Tony I think discovered was that the sensors are there the sensory networks are there um but they terminate the ship's Bridge so the data doesn't escape the The Vessel typically um what we're doing is we're tapping into those those pre-existing um sensor those buses um to to take a stream the sensors are looking at things like engine temperature uh simple things like engine RPM engine load torque and then into much more um minute detail like uh the the pressure in the turbocharger Inlet um exhaust pressures uh on individual cylinders uh cylinder temperatures I mean you know so some of the vessels that we work with have up to 2,000 sensors on board so they are tracking pretty much everything okay okay so with 2,000 plus sensors potentially on board that's a huge amount of data and with the increased frequency that you're capturing that over what it used be you know how much data do you gather in a typical day off a typical vessel for example that's coming back to yourselves um typically somewhere in the region of 3 to 400 Meg um but we've spent a lot of time both on the edge on the on the device on board the vessel and uh working with yourselves to to try and find ways to to compress that down so um it means not capturing duplicate data very very difficult actually and a lot more challenging than we first anticipated um and then just building our data structures um in a way that allows for massive massive compression um so we spent um a lot of time working with you know your engineers and Consultants at mongod Deb to to find a structure that works uh we're now at a point whereby uh 300 Meg will compress down at a stored data size to honestly quite literally three Meg um as it scales so we're getting some really really good uh scaling on on the data but in terms of transmitted data um you know raw data would probably be yeah 3 to 400 megabytes uh we could capture gigabytes maybe even a terabyte a day um if we really wanted to okay so look we're talking about large scale data but with lots of clever tricks Etc getting that it's own something more manageable before we dive deeper into what you do with that data tell us a little bit about the the tech stack overall B at a high level you know how things work yeah so uh fundamentally I guess we break it down into three main sections maybe four main sections so first of all it's the hardware so that's uh The Edge side uh compute a lot of filtering taking place they're taking um Network high frequency network data uh encoding it working out which uh fundamentally which sort of protocol it's being transmitted on and then digesting it and transmitting it in a way that works and is suitable and is consistent for us so that's the hardware side of things then we have I guess the the UI um which is the the web stack so uh a user dashboard that's built in um in nu uh and then a query engine um which is a a very powerful API that we've built out so it's it's more than just simply an API because uh given the way that we have um multiple sensors across multiple vessels um with uh specific differentiators we actually first we have to generate the query so understand uh and translate what that query looks like and then actually go and query the database um so that's kind of the web stack and then we've got our um analytics engine and our data engine so our data engine ingests uh thirdparty data any data that we're sort of capturing from third parties that isn't generated by our own um our own sensory devices and our own on the vessels and then we have the analytics engine which is where we yeah where we carry out all of our ml okay so I mean you brought up ML there um I wanted to you know obviously dive deep into the AI and why AI um and I suppose at a high level from my limited understanding of what you're doing this is this is is about trying to you know predict the potential catastrophic failures before they happen as you mentioned Tony before you know the The Vessel grinds to a halt in your experience to date you know how far in advance are you able to be alert to a potential problem that here to for wouldn't have been available in terms of insights to anybody on the vessel until that siren or light or gauge goes Ary yeah I mean so you hit the nail on the head there with the the siren sounding um or alarm sounding and and traditionally that's what happens is you you get um you get an alarm triggered uh based on you know predetermined thresholds whereas if a temperature say for example um a Turbocharger Outlet temperature exceeds 500° then you know there'd be a notification provided to the engineering team and they would make their way to the engine room and essentially assess the problem if it's determined that maintenance needs to be conducted at that point in time then they they they would shut everything down and and conduct the maintenance um obviously there's a couple of tricks and things like that that Engineers can can do to to keep things going for a little bit longer but you have that tradeoff then of you know continuing to run something in a detrimental condition um or stopping the vessel conducting the maintenance the the trade-off there is from a commercial perspective if you stop the vessel okay you may lose eight hours by conducting that maintenance but if you continue to run that piece of equipment it may cost you 90,000 100,000 $200,000 to to actually repair it destination Port purely because you pushed it Beyond its limits and there's um you know there's additional maintenance and additional damage being being done so that that's that's how things kind of planed hand out up to now now we see um obviously more real-time tracking of data uh the ability to to assess in real time and understand exactly what's going on but for the Shide team to be able to support those um Engineers uh and the crew me the other crew members on board as well in in the decisions that they're taking around maintenance um you know waiting up the factors from a commercial standpoint because with any business or operation there's commercial pressure to get caros from from one port to its destination Port um but yeah in terms of prediction fact that in the old ways like I say you've got a notification via an alarm um traditionally you would take a pen and a piece of paper and you would track those figures over a number of days and understand what the trend looked like now you can Trend it over a couple of hours um and we're working as as much as we possibly can to extend that prediction window uh which which will allow crew members to intervene at the the most convenient time and the most optimal time Point excellent so let's get technical now Ben if we can so AI it's in the name it's it's there this is how you know this is your secret sauce how are you leveraging Ai and obviously you needed a you know training there as well too how did you go and train that Model H to help you do what you do in your solution yeah so it's probably best if we start just explaining a little bit about the the data Pipeline and how we um how we capture the data up front and and then what we do with it and how we apply it sure um so I think we sort of came at this because it was kind of the excuse the pun a huge ocean of data we needed some kind of honestly the the number of times we'll just in the office and we'll just realize how many Maritime puns are just thrown into day-to-day life it's crazy so um yeah if if I do end up dropping a few more I apologize but problem a huge ocean of data uh we needed a view it um I needed to take I guess a framework a paradigm um in which to view the data so the the first point um on on the pipeline is uh what we call V or um vessel activity recognition and that is basically just taking pretty crude metrics to understand what the vessel is doing so little bit of very simple um supervised learning on that to to establish a uh a Class A classification model so that says the vessel is underway the vessel is at anchor the vessel is docked birth that port or or whatever it happens to be so uh with that we understand voyages so and and the voyage then is simply a time frame on which we then want to run more analytic more models so take time frame uh or or that window for the for the voyage and then we ran or or run ran uh a degree of supervised learning on that model um to to to help predict some anomalies so we take a window a time period and then we just run anomaly detection on that and that's pretty straightforward anomaly so um using uh isolation Forest as sort of the the main model that we've used for that one of the SK lean models it's quite straightforward just detecting anomalies in the data sets and that's relatively relatively easy to do once you have the data and can those anomalies be obviously look again not familiar with vessels and large Maritime things Etc more familiar with with btes and data and coding but obviously most vessels unless it's the same make or type or model or size whatever you you mentioned that kind of supervised learning um you know in doc on a voyage etc those anomalies do you have to monitor for a period of time to be able to filter those out obviously then to to get a profile of that vessel perhaps right yes that's correct so this is where you start getting into like crazy multi-dimensional stuff but on a single vessel yeah you can um we use typically we use a 30-day um on period to just understand the Baseline for the vessel um and then any anomalies are sort of Taken Beyond uh outside of that window fundamentally uh but even still there are anomalies that you can detect within that window that the vessel might not be that the the crew might not be aware of so um certain as I mentioned earlier certain cylinder temperature sensors running slightly high uh is indicative of of certain things um and it's those kind of behaviors that we look for or those kind of markers that we look for in the anomaly detection so we have a huge data set or a window of data and then within that within that time period we're looking at for certain markers okay okay so within that then you're you're obviously you mentioned obviously you know you've got your query engine your your API your data analytics your data engine um following on from that how how are you leveraging I suppose mongodb and maybe before that why did you even choose Monga to be as part of your Tech stack in the first place yeah so um I suppose there's there's probably then actually beyond that point a couple more places in the pipeline where we where we're using mongodb um from there actually we then enter a chunk more supervised learning which is where we use the likes of Tony's years of experience and Tony's colleagues and and te's experience to understand the root causes of those anomalies so we're then trying and and that's what we're working on at the moment is building a data set of effectively a labeled data set at this so all of that data is in mongodb um as well and we're using data L to pull that out into the lacks of data breaks um and other uh Cloud uh resources basically so uh just in just on the on the mongod D bit itself and why we chose yourselves I suppose a couple of things one uh time series data super important and obviously uh you guys are investing a lot of time now in Time series stuff which is really really cool um uh and then secondly was we were looking for a very sensible um approach to compr ression because the big one that the the most nervous my my biggest fear in in everything that we were doing was just the Reams and rooms of data that we're going to be collecting uh obviously we want something that's fast fast to access quick access um I think Tony you might just have a little bit of feedback going there at your side yeah I think you popped out his headphones and he's trying to put them away in the case that's what we're hearing no worries Tony um so yeah so uh gor size was I've got a load of feedback now okay these things happen no worries I think Tony when you took out your airpod getting this is it gone think we're good I think we're good Tony's gone no worries these things happen it's live back we drag you back good man perfect sorry so so Ben you were you were talking about compression there yeah before Tony took out his headphones and messed everything up on us uh yeah exactly um yeah time series data and compression size so time series for fast reads and wres well fast reads in particular and then uh and uh yeah the compression as well that comes with that so that was fundamentally it but also um I think from like a a high level um uh data structure perspective we don't know what data we're going to be working with a new vessel um could have a very different sensory infrastructure and network than the any previous vessels so we we needed to operate in kind of a schem environment if that makes sense and again also being a startup being quite young we want the flexibility to just be able to go in Any Which Way Direction without having to worry about schemas necessarily or migrations or anything like this so it's it's kind of one big win is the sort of the nosql non-relational element to it that was actually what pointed me here or sort of towards the um non-relational direction to start with and then from there it sort of progressed into into checking out who uh well checking out and understanding your your your abilities so yeah and and by the way like I know I'm smiling about it it's because I'm a big fan like I'm a huge huge mongod DB advocate so excellent but listen of course that's obviously music to our ears but I think you what you say there about your startup you don't know the structure of your data you don't know your schema and that's really what you know our document model allows you that superb flexibility so that as you progress as you manage to grow and change in terms of what you need to host and to analyze you know the the document model is superb that we see that time to time all of the time when we speak with startups in particular you know because they're not you know building a an MVP and then having to totally refactor it again further down the line Etc this you know that everything can grow with that the other thing that you mentioned which I think is great to hear because I I you know the time series essentially the we launched Mong for time series data probably two years ago maybe a bit more than that as well too and that was key um in opening up a lot of the iot space to us which obviously you guys exist in um because you know that data as you said it's a it's an ocean right it quickly builds up and you need to manage that data in a certain different way than just writing to your nearest clusters and filling that up and and then you're you're worried about the structure of that data going forward so for those of our audience who are not familiar with that just search for Mong to be time series especially if you are in the iot space because uh it's a particular bonus I think that we managed to get again again from our structure from our document model allowed us to to straightforward take care take care of that but it certainly does allow you to it and build as you say faster and quicker as well too in that space um so it's great to hear um you know the the reasons and and the background as to why you chose manga to be I suppose you mentioned it at the beginning Ben and Tony don't worry I'll grab you back in in a minute I'm not leaving you out of this conversation um but you mentioned at the beginning that you know mongodb were helpful I know you guys were on the startup program uh along with mongodb how did you find find that or was was that more on Tony's side or your side Ben I don't mind who answers the that particular question no I'll I'll go if uh yeah turn skp mic on so I've I'd used yourselves before previously um so I came to to you guys actually looking for some kind of um startup program um in in terms of the the benefits uh you know I think the main one is just being able to forge a good relationship with uh um your the sort of uh the the engineering the consultant engineers and um for advice and to understand the direction in which to take things early doors um I know like we've just mentioned nothing really that we do is is irreversible given the document structure but it is just nice to know that you're sort of going down a path that will last you for you know the next 18 months 24 months without having to worry too much about um you know the a quick rebuild all that kind of thing so yeah getting the support from from your guys uh very very helpful um and then I also think you know just in terms of the the exposure and credibility it's helped give us as well um it's it's been really useful uh you know events like this uh some of the the Press coverage that we've received from yourselves as well as a startup like you'll take you'll take everything you can get um so it's you know it's probably those two things I think not to mention the the free credits are helpful as well always good always good well look I think you've clearly outlined I suppose why you chose mongodb and and and the benefits that that brought to you and you know the various products that we had that helped you let me talk a little bit about the challenges Ben or Tony you know what challenges did you encounter whilst developing your platform and your system either from a technical point of view or even at your space Tony perhaps maybe convincing some vessels to go with you and run with you Etc as well yeah I mean um from the commercial side of things the challenges are huge as you can imagine with any business but just stop you for a sec Tony do you have can you come closer to whatever your microphone thing you might have on your desktop or somewhere there okay hopefully it's a bit more clear it's not bad don't know what our audience maybe might let us know but did your airpods die on you they did yeah uh they certainly did one airpod died and then it was just downhill from there um we'll try our best we'll try our best but from a from a commercial standpoint you know the biggest challenges are um actually working with some of the boners to get the hardware element of our product on B the vessels so if you're unfamiliar with Maritime it's full of very skeptical people they've been burnt in the uh in the past on you know new technologies and Bad actors in the past very robotic it's not that bad but yeah um so you know people in the past have been burned and building up that trust to work with a small company can be a hurdle to uh to overcome but uh yeah as the comment flooding saying the audio is UN click your mute button for a minute or two there Tony and and so Ben's gonna take the the next slot of a few questions so Ben hope you're up to this I I think the other thing too and I know we talked about a little bit is obviously you know this as you said ocean of data heading into a data Lake Etc scalability and I suppose security and privacy talk to us a little bit to the audience about you know how Sito manage that and and kind of how important is that within the martime industry yeah I mean I I think um what a lot of people probably don't appreciate is the um a lot of the once you get into a networking uh sort of environment on a vessel or any um any I guess heavy industry or even uh car transport that Network which which you are tapping into is you know is often able to write to the to the sensors and controllers on within the within the uh within that equipment so uh first of all first and foremost um we have at a hardware level made sure that we are a readon device so okay we can only sniff Network traffic we can only read there is no physical Way Beyond just a software switch but a Hardware switch that you can actually transmit data uh to that Network so there's no way to hijack The Vessel or take over what's going on or cause the to do anything it shouldn't be doing because it's it's read only correct correct so and that's at a hardware level obviously we we back it up with software but uh fundamentally it's it's Hardware safe uh so that's that's thing one um probably the second thing is obviously endtoend encryption on all transmissions so everything uh that we transmit from The Vessel uh to the cloud is encrypted there and then actually from a mongodb standpoint uh any data that would be um that would help identify a vessel so um for instance it's location which is publicly available uh its name um it's it's International Maritime organization number uh these data points are all encrypted at a driver level so uh everything stored in that regard in the database um is encrypted as well so and then of course we we carry out pen testing uh standard pen testing just to ensure that we are uh safe and compliant there and then also because of some of the contracts that we have coming up we are uh working well we are almost uh ISO certified and uh have UK cyber Essentials security as well so okay because I was going to go there I would imagine yes totally on top of data security and privacy it's a very regulated industry with probably tons and tons of compliance measures Etc so those are are something that you all have you have to get out of the way as well too with your with your technical solution and I suppose going back to the fact that historically this was done all on board a vessel is there was there any anything Key you mentioned the the satellite Network earlier was there anything key that was a challenge or a barrier to be able to offload this data back to back to the database and back to your your dashboards Etc yeah I mean connectivity is the the big bottleneck and I think it's probably the reason that the industry hasn't really got the sort of uh technology uh in place that other Industries have um and I think if Tony was able to speak clearly he'd probably talk now about starlink and how Okay in in the same way that you know the iPhone opened up for instance Facebook um and the internet Amazon back in the in the late 90s um we kind of feel that starlink and and a lot of these Leo um uh constellation connections will enable the industry to actually start moving forward um because connectivity currently if you if you don't have a a starlink connection you are probably on a you know a one megabit connection at best okay with very very limited um uh uh bandwidth and and high high latency so that's a challenge in its own right uh kind of feel that we are positioned perfectly to sort of piggy back on the on the infrastructure wave that's that's coming yeah so it's it's a kind of a a lift you're going to get because those technical infrastructure barriers should be disappearing or dissipated at this point in time I know you earlier talked about the compression of data you seem to be doing a super good job there but regardless you still need to do things faster and quicker and more reliable so those networks and I'd forgotten about starlink I assume yeah they do have a roaman starlink now it used to be that starlink had to be stuck at a certain address Etc but it works for maritime now too yeah and I mean the the big thing there is the cost reduction so an existing vsat vsat being the the typical uh internet provider for for maritime uh typical vat connection is super super expensive uh you know we're talking thousands of pounds a month uh multiple thousands uh star link now on the the Marine package I think you you're down maybe as cheap as 500 to 1,000 pound a month um for uh you know at times a gigabit connection so um yeah it's it's really opening doors okay I'm gonna try Tony once again I have a question for you Tony on this so we' spent a lot of time talking with Ben obviously about the the structure and the tech stack and the ingestion and and analytics of the data in what you've deployed so far in the various vessels you know how how quickly did did those vessels see a return on investment on this how you know I suppose you're almost in this insurance game you know if everything goes well and everything's working well ideally you don't want any of it's not a red light in your dashboard I would imagine it's a it's an alert or a chart or something but if everything's going well with the vessel you guys are not invoked but if it isn't going back to the AI and the insights how quickly do you get the prior knowledge and any examples of were you know as you said earlier it might be 30,000 pounds a day if a if a boat is stopped because of a malfunction any examples of where by quite quickly within a roll out of Sito you've helped a vessel in that respect and therefore they recoup obviously maybe the the costs of your service and your product yeah yeah well hopefully you can hear me a bit clearer yeah that's much better than it was a few minutes ago okay change the microphone um excellent but yeah just just on that so you know we we um we recently undertook an analysis project uh with a a very large ship owner um and they provided us just with a sample set of data so we can actually you know ingest third party data as well although higher frequencies than what we traditionally see from our proprietary device um sorry lower frequencies than what we see from our propriety as as I mentioned earlier on minute interval whereas where every second um but you know from from analyzing just a quick snapshot of four four and a half months of their data we immediately were able to identify an issue with one of their cylinders um and managed to narrow it down you know from first analysis okay cylinder 3 on the main engine is operating um 12% higher in terms of temperature compared to the rest of of of the cylinders within that engine um from that we then gave a weighted factor of what the likely outcome would have been so you know it was it was sort of distributed across three different elements um one was uh suit buildup on the exhaust gas valve uh one was um fuel injectors not actually atomizing the fuel and dripping excess fuel into the combustion cycle uh and then the third one was around fuel rack position and whether something had um had had an effect on that fuel rack position which would then result in more fuel being injected into the engine than was required so in terms of you know an return on investment that was almost immediate based on just an initial sample um it it essentially proved out that the models are identifying the anomalies which we're expecting them to identify um and and providing value straight back to that customer to say okay there's an issue here maintenance needs to be conducted immediately in order to to prevent further issues down the line and you know those further issues could could res that means is you get a buildup of of soot within that com uh within that cylinder liner um you have a scavenge space which essentially feeds air into the combustion cycle and um that soot then builds up within that space you've got hot air you've got uh High pressures you get combustion and you ultimately will end up with a scaven space fire or you know worse so we've we've identified the risk we've mitigated that risk but we've also given them the information that they need to be able to act upon that risk uh but the mitigation mitigating factors in place and stop something from happening which would ultimately cost them from a commercial standpoint so yeah to to wrap that all up quite quickly is the answer uh in terms of to me it's it's you know you you answered it even more than quickly you did all of this on Sample data um prior to you know being on board in the vessel properly and therefore your proof of concept even before a deployment was already saving money right so I think that's that's super strong the when did you deploy first I got should have asked you this back in the intros how long has Sito been around when did you first get your deployment and if we could look forward a little bit to the Future what types of new features products Etc are you thinking of or what new I suppose uses of AI as well too so that's a a multimodal question so when did you deploy first and what's in the future what's around the corner for you guys yeah so first deployment was on um two small vessels uh at the local Port um and that was back in back in the end of 2022 um so we built we built an MVP um and the Marine manager down there to to essentially run through what a trial would look like and what the value would be for him um so got the solution deployed monitored over a period of time uh and yeah obviously that that went well and they turned into a you know commercial partner um you know happily paying for for a solution where where they're getting value on a daily basis um in terms of moving forward and what we look like in the future you know so we'll be at um at 23 vessels by the end of this quarter uh live and they range from you know your small Harbor pilot vessels tugboats through to Offshore Supply uh as well as very large container ships and bul carriers um so quite a mix but a welcomed mix let's let's say you know covering off as many uh different classes of vessel as soon as possible is is is really good for us um it gives us that exposure to the rest of the market so yeah that that's that's where we're looking um by the end of this quarter by the end of the year we're expecting to be up closer towards 70 80 vessel Mark um everything going well with with the the the business that we've got currently underway um okay and then I guess to go on to your next question about products what's new what's next so next week we're actually launching and and you're hearing it first here because we haven't put any press out about it yet or anything like that so well done Shane on getting the exclusive I suppose um you heard it here first uh so we're we're launching a product to help with with EU ETS compliance so EU ETS is the European Union's emission trading scheme and from the 1 of January this year Maritime now fell under that directive so what it means is for every ton of CO2 produced the um the ship owner needs to surrender one EU allowance so just to put that into perspective for every ton of heavy fuel oil burnt there's about 3.2 tons of CO2 um so you can imagine that needs to be monitored and tracked quite quite carefully those euas need to be surrendered otherwise there's penalties that are you know imposed against the ship bers so as of next week we will be launching um a SAS model on that product it will allow ship owners to easily quickly sign up it doesn't interrupt their existing workflows we essentially just send over a document uh that that's sent out every day at noon to a designated inbox we pass that information and then analyze it and produce their um their compliance report all wrapped up into our existing dashboard uh and then from that you know in the in the next month we'll be deploying a feature which which looks at predicting what future eua requirements look like um to help you know vessel owners operators understand their eua visibility and and what the Outlook um would be what the cost would be if they were to take alternative routes between any port and an EU Port okay excellent so you're branching out and uh best of luck with the the the launches there going back to the AI space we often hear or the the the media and the News will will fearmonger AI is stealing all our jobs if I was been super cynical you know there's a very an engineer comfortably sitting on a vessel somewhere today going I'm important because I get to monitor these lights and these gauges Etc is AI going to take that engineer's job is ISO gonna take that engineer's job no absolutely not um what will'll do for the engineers rule it will make it easier it will enhance their you know their ability to perform their job giving them information um you know highlighting areas which are of concern ahead of time so things can be prepared on board spare parts can be ordered uh ahead of time maintenance can be conducted in you know the most favorable time periods so rather than having to as I touched on earlier wake up or do a full day's work then go back down and have to do another 12 hours um you know working all the way through the night Engineers will be able to schedule and and and conduct the maintenance that they need to do in time Windows where it's more normal operating hours helping out with things like the hours of work and rest which are dictated by the maritime labor convention and really just enhancing the welfare as well as the the ease of the job um for the engineers but by no means will will AI take the jobs of Engineers and plus you need somebody with a pair of hands to to go down and and actually conduct that maintenance themselves I don't think uh an artificial intelligence model is going to be able to pick up a SP anytime soon yeah the robot overlords are not coming just yet anyway well answered Tony well answered I think you know that that's key it's it's about providing them the information so that they can be proactive and and understand and plan Etc as well too going back to you mentioned your first deployments in in 2022 the end of 2022 um obviously the end of 22 coincided with the mainstream understanding of AI and large language models and foundational models Etc as well too Ben how do you keep up with what's going on in the AI space it just seems every time you you know you think you've got a grasp or a handle of what's going on there something new comes out something bigger something better something Bolder something promising more uh insights more value more knowledge how do you as a development team keep ahead of what's happening in that space yeah so I think there's probably two answers to that question so I think all of the llms that we are seeing you know grabbing the headlines uh and that have been over the last sort of 18 months um they are all fundamentally just transformers so the there there hasn't been albeit there has been significant iterations on a an original design I think uh the original attention is all you need was published in 2016 or 2017 it's just been a case of f tuning or enhancing that existing model so there hasn't actually been that much uh more of a of a um concept if you like to sort of wrap your head around to understand what's gone on in the space uh at a at a deeply technical level um so that's the first part uh the first answer I guess uh and then secondly it's it's honestly I'm sort of ear to the ground on on Tech Twitter um and and in any as we need to call yeah right sorry X sorry yes uh in in Tech X um ear to the ground there and and then just generally involved in in Tech uh ecosystems so um on a number of Discord servers uh I've got my uh WhatsApp groups with fellow devs and we'll we'll ping uh various bits and pieces across to each other when we see something of Interest so uh just trying to stay involved fundamentally at a social level and then from there you can sort of go down and deep dive into the into the technical Nuance of it all so it it's it's no secret sauce it's more keeping a breast at what's going on using your peer networks you know trying you probably I would imagine are experimenting all of the time um in terms of what can be applied to pedo uh Etc too I think in the Journey of building and this be answered by either of you what are the lessons valuable lessons that you've learned in in building this relatively young company into what is you know a very very very old industry Maritime has been around for you know thousands of years at this point in time and what have you gained in terms of knowledge about bringing AI to a very very historically kind of you know boats have been the same shape for God knows how long there's been an engine in a boat or a sail in a boat but fundamentally they're the same bringing AI into that I suppose probably somewhat maybe old-fashioned I know how my boat sounds I can hear if the engine's going to cause trouble before any of your fabulous technology AI intelligence is going to let me know any any learnings there maybe Tony or Ben yeah I mean um when you approach the market and you bring new technology and new ways of thinking um there's a a process of Education or a process of learning for you know some of those companies those who aren't maybe is Forward Thinking um overcoming that hurdle in the first instance before you can actually sit down and have the real um you know commercial conversations that that's that's the biggest step because as as I mentioned try to mention before the uh the mic gave away earlier on people are are very skeptical of of new technology you know there's there's been companies in the past where where they've been promised the Earth and and never delivered so people have a guard up you need to overcome that guard you need to overcome the process of education and once you break those barriers down that's the point where you can really start to create value together and you know we we take the approach of when we're not a c when we're not a supplier and you're not a customer we're Partners in this we collaborate we develop we build the future of of what Maritime looks like um but yeah I think in terms of the the hardest challenge is it's it's that education process some people have only just learned to use computers or um you know particularly the older Generation Um I know and my mom's trying to use a phone it's like it's uh quite quite a sight to see but you know are in the 60s and 70s in some cases and they're not as literate with with technology educating those decision makers can be a process to overcome and to tag on to that I think um it's from where I'm sat someone that doesn't know Maritime at all uh it's abundantly clear how important it is to have someone who does know the industry at the front of the company I think uh you'll often see lots of tech companies try and break into an industry that they know uh some or a little about um and without having a figurehead who deeply understands the problem that you're facing uh certainly in in in Maritime I think we' we'd struggle to have have got to over have yeah well you certainly yeah Tony you've been there and done that and had the sleeves rolled up Etc where do you see I suppose looking forward and and Beyond Sito and beyond your own area of expertise and and and your own products where do you see AI in the maritime industry as a whole how do you see you know that changing in the future how does how does Maritime evolve you mentioned obviously emissions and we obviously you know that's the next thing that you're going into but you know Ben you touched on starlink and connectivity you know what do you see coming to that industry like are we going to get to the point where the the robot vessels are just doing their own thing connected back to seos and they can pull into a port if there's something wrong I I think you will get to a level of autonomy um you won't replace crew I would say not in the next 20 years and that's just the the way that legislation within the industry adapts and changes um you know the maritime industry supports millions of jobs it's a 14 trillion dollar a year you know economy of its own essentially um it the entire Globe relies on it um I I I do think you will reach a level of autonomy it won't be on the the linest the sort of large scale uh ocean liner scale but maybe on those smaller Coastal faeries the short sea routs you'll probably see it there and I think where we play into that is is being able to provide the oversight and the understanding for The Operators of those vessels to when mainten maintenance to be done um as well as you know helping people within the insurance sector shoulder the risk of operations with those vessels or and even operation in the short term of of the more traditional Maritime assets which are out there but a level of like I say a level of autonomy will eventually arrive um I don't foresee that being for at least 15 20 years purely down to legislation changes okay you know that's very understandable I think we've been saying the same thing about aircraft for a long time you know the autopilot does most of the work but you still have two people sitting up the front of that plane um as well too is there anything within you know where you are currently you mentioned you know upwards of 2,000 sensors already on any vessel Etc is there any you know piece of the I suppose the data the information that's on board a vessel that Sito can't touch or you can't get at or you'd like to get at to add to your product portfolio is there anything else that you can get involved in in the you know the awareness of that vessel in transit or in Port I mean if we look at it from a risk standpoint um an understanding of the cargo itself would be of great value um particularly in the containership space you know you've got um in some instances Undeclared cargo which is potentially hazardous uh you know there could be batteries there could be pyrate Technics and and and various other things and the you know being able to to get handle on on declared cargo or understand exactly the condition of container maybe something that we look at in the future um but right now from a systems at a systems level we essentially tap into to everything that's connected to that Alarma monitoring system as well as all of the bridge equipment so we're in a very good place from the vessel's perspective moving into the cargo space is something that you know will be useful to understand um and then you know next steps for us moving away slightly from Maritime would would be more towards the renewable space as well um with with offshore wind turbines yeah it's an interesting point actually because I suppose you know given what you've outlined Ben given your solution given the platform you can analyze data from pretty much anything that has a sensor attached to it you can ingest that and you know use your supervised learning and your learning and all of I suppose actually it leads me to one other question Ben as you've gathered all of this data over time first of all how useful is historical data to your platform and to your system and then secondly the two-parter for this is what do you do with that historical data do you offload it at some point to a data lake or something else as well to Federated storage yeah that's right so um first question uh historic data very important uh because it allows us to to you know on a vessel by vessel basis understand uh baselines we can then obviously do time series analysis on the uh you know any anomalies uh against a previous pattern or previous Voyage so yeah super important in that regard um uh at the end of uh the plan as it currently stands is uh when we tip over the the 12 month of stored data um to wrap it up either into yeah some into one of your uh Federated data platforms or um wrap it up effectively into into blob which I know that you guys can uh still access as well we haven't actually got to that point on any of the the uh the mongod DB data that we've got so far so uh it's probably another three or four months away until we'll have to cross that bridge but that's that's the plan okay okay well look we we can certainly help you out there as well to there's there's plenty of case studies and what we've done with other clients with huge amounts of data such as yourselves I know we've gone over our hour I think this has been superbly interesting conversation as somebody pointed out on the comments I came to YouTube to learn mongodb and look what I found I mean I think you've we we've seem to have maintained our audience throughout this and and even though we're not putting code on screen as we usually do we're running through a demo and and that's because of the sensitive nature of some of the information that you are gathering that's live in your system I think the fact that we've maintained our audience is superb and and speaks to the the interest people have in this level of gathering ingesting analyzing performing metrics and analytics on large large data sets so thank you so much any final questions before we leave our conversation here either from you Tony or for Ben any final insights or or anything that I forgot to ask you as we went through this conversation that you'd like to bring up oh well that's a good question um let me have a little thing I I mean the only thing that I would probably do is shamelessly plug us on at the end of this and just say you know if any of the audience do have a link into a shipping company and they are facing engine reliability issues then by all means Reach Out find a contact form at our website and um yeah do get in touch we can we can certainly help you yeah if any of you have a spare cargo ship lying around that's just not working quite right and you you want to get some insights yeah hit Tony or B up you know throw deploy deploy it there speaking of deploying anybody joining into the show and wants to you know learn more and and get involved and isn't already using Atlas by all means go to to one of our short links there all of the links that we have are MDB do link so for this episode CC Atlas 22 gets you straight through to the registration page anybody wanting to know a little bit more um can also see a case study we did with with cedo and a couple of other AI companies that's published on the on the mongodb website so you can certainly check that out um there's lots to read there and we have a lot of AI startups and AI guests on cloud connect so just make sure to to keep an eye on what's going on um you know in future episodes the Lynch pin for mongodb obviously for any of the AI things is is our Vector search we launched it at local New York last last year in June um it's generally available now you can go in and use it and that works as you know Ben alluded to it works with Mo all of the gen products from our partners most of the large language models that are out there you can use your own embeddings and you use essentially Atlas as your vector store uh which is ideal if your data is already within Atlas because you're storing your vectors adjacent to your data you're not spinning up anything extra you're not doing any of ETL loads you're not extracting transforming uh your data it's all back in Atlas so that's my part of the plugs that are over now most definitely um this has been very very interesting I I'm amazed at the journey that you both have been on in quite a very short space of time I'm amazed at the fact that you know this year how many like do you know Tony or B how many vessels like what's your total addressable Market how many how many potential vessels in the world could you have cedo deployed on you you mentioned you're on 22 which is superb for a young company a young startup and from all different sizes of vessels but obviously you know this can be applied to any any vessel that's in the world today yeah so um globally there's around about 120,000 commercial vessels registered with IMO numbers um you know we started the the approach to the market with a an ideal customer profile targeting vessels which between their their first special survey and 14 years old which will 15 years old is the Third special survey uh those windows um traditionally on short sea voyages because we can then dual have dual connectivity with starlink and um an LTE connection 4G 5G um to rely on you know the most efficient and and fastest internet speeds um and and really just you know ship owners or ship operators that are Forward Thinking in their approach they they've either collected data in the past or they they understand that data is going to unlock a lot of value for their business in the future um and and we put that at around the 900 vessel Mark um based on the research that we had so that's our our initial ICP moving beyond that you know we we'll be aiming to scale up to to three four 5,000 vessels um in the future but right now let's let let's keep our feet firmly planted on the ground and just keep plugging away um you know building through the motions and onboarding as many people as we can uh in a sustainable manner excellent I mean it's great to hear those Ambitions and hopefully you know obviously Mong will be with you all the way there as well too I know you raised some funding uh last September or so or before that as well too that will allow you obviously to grow this journey and grow your company yeah so that was um that was a one and a half million pound seed round that was led by Howen Ventures so Howen are an insurance broker they've recently launched the Venture fund and we were actually their first investment from that Venture fund um and with that that gives us a a huge distribution Network as well um they've got a team of 50 Marine Brokers distributed across the world uh all with access to ship owners and and as as their clients so if we can you know add additional value through those routs then then that's fantastic it's a win-win win situation for for all parties involved um but yeah other you know other participants in that round we have a an amazing group of investors who were first Incorporated they followed on um as well as uh Founders Factory who are in a joint venture with um Aviva another insurance company so there's a lot of interest from a risk mitigation standpoint um and that's why we we find that insurance companies are supporting this as well excellent listen that's very very impressive to hear and you know raising money is a job in itself let alone trying to run the company from a day-to-day business and and roll out new products and get new customers so so many congrats on that and I hope it gives you enough Runway to to get to the next big launch and big roll out for for stto and um I think this has been a fabulous story I think one of the things that comes out and Ben you really talked about this was the help that mongodb can give to Young startups so just throwing in the link up there for mb.com startups we have a startups program in fact we have a particular strand which is the AI startups program as well too so if you if you click there you can provide your details and get registered and obviously we'll reach out to you H to see how we can help how we can get involved as well too and but for now Tony Ben this has been a fabulous conversation I do appreciate your time I'm thrilled even more now that we managed to get this episode on air after a couple of false starts um but thank you so much I certainly will keep an eye on cedo's progress and um hopefully we maybe get you back in you know another 8 12 months time and see how you're doing but for now thank you so much for for joining us no thank you very much for having us and uh yeah we look forward to that excellent excellent and to all who joined us on LinkedIn and YouTube we do appreciate that thank you so much for all the various comments that were coming in as well too I didn't really get to click on them all but I do appreciate that and the fact that everybody's joined from different parts of the world is always a super surprise we stream Cloud connect most Thursdays next Thursday again we have another show with a company called super duper DB again in the AI space so do join in for that 12 eastern US 5 PM UK 6 p.m emaa and and elsewhere so you can you can do the math on your own time zones but yeah once again Tony Ben thank you so much and to everyone who joined us we really appreciate you we hope to welcome you AB board the cloud connect live stream again soon do take care