The Shifts Project for world-class AI

Never heard of distributional shift? If you’re in shipping, you should read our post on what it is and why you need to know!

Put simply – it’s one of the greatest challenges to effectively harnessing data, including the use of artificial intelligence, in every sector – and especially so in shipping. Handling this data phenomenon elegantly is the key to unlocking major efficiency improvements on your ships. 

Over the past five years of research we have become industry leaders in this specialty – allowing us to make an average 8% efficiency improvement per vessel through performance routing powered by AI.

But we know that, together, we can increase these numbers – and that collaborating to share approaches to handling distributional shift can not only make shipping greener, but also help sufferers of Multiple Sclerosis, help accelerate the autonomous car revolution, and help improve the lives of millions.

DeepSea and the Shifts Project

To that end, we have launched The Shifts Project alongside a group of researchers from the Universities of Basel, Cambridge, HES-SO Valais and Lausanne. This is an international collaboration dedicated to studying and improving approaches to the critical topic of distributional shift in high-stakes, real-world settings (such as shipping).

“This exciting collaboration will help increase the real-world impact of data-driven approaches such as DeepSea’s. Importantly, it will also provide a benchmark for shipping companies to understand where the industry is in terms of AI – and the sort of progress they should be expecting from their collaborators.

Konstantinos Kyriakopoulos, DeepSea CEO

Test your (virtual) metal

On that note, the Shifts Project has also launched the Shifts Challenge, an open initiative which encourages AI researchers – industrial and academic – in the shipping domain to test their own models’ capacity to handle real-world shipping challenges. The winners and runners-up will receive awards and – above all – a serious “stamp of approval” from some of the world’s top experts in the field.

More information on the Shifts Challenge can be found on the Shifts website.

Working with AWS

The Shifts Project is powered by AWS, providing a fully scaleable environment for evaluating the submitted models. DeepSea's cloud-based infrastructure is also powered by AWS end-to-end - including platform hosting, data storage, and a complete AI environment.

Wallenius Wilhelmsen moves to fully AI-based voyage planning with DeepSea

September 7th, 2022: Wallenius Wilhelmsen is becoming the first global shipping company to adopt a fully AI-based approach to voyage optimisation across its entire fleet of 120+ ships. The company will roll out DeepSea’s Performance Routing software, which provides vessel-specific route and speed plans, in the final quarter of 2022 and 2023. One of the most forward-thinking companies in the industry, Wallenius Wilhelmsen has set ambitious targets to reduce emission by 27.5 per cent by 2030. This work with DeepSea is an important step towards meeting them. Geir Fagerheim, SVP Marine Operations at Wallenius Wilhelmsen, says:
“No human being, no matter how many years of experience they have, can compete with these automated sailing instructions. It reduces emissions, it reduces fuel consumption, and it increases safety during operation. It is a win-win in all aspects of sailing.”
This landmark decision, the first of its kind globally, was not made quickly – but followed 18 months of rigorous step-by-step testing. The numbers that ultimately came out of this comprehensive trial period are significant: A 6.9% improvement in vessel efficiency and more than 170,000 tonne predicted reduction in emissions across the fleet. However, equally important to focus on are the key learnings that emerged from the 18-month period of intense collaboration that led to this partnership. On the 13th October, join Wallenius Wilhelmsen and DeepSea virtually, as we lift the hood on this exciting validation period and together discuss our key learnings for decarbonisation with voyage optimisation.
“It reduces emissions, it reduces fuel consumption, and it increases safety during operation. It is a win-win in all aspects of sailing.”

Launching the Shifts Project for world-class AI

Never heard of distributional shift? If you’re in shipping, you should read our post on what it is and why you need to know!

Put simply – it’s one of the greatest challenges to effectively harnessing data, including the use of artificial intelligence, in every sector – and especially so in shipping. Handling this data phenomenon elegantly is the key to unlocking major efficiency improvements on your ships. 

Over the past five years of research we have become industry leaders in this specialty – allowing us to make an average 8% efficiency improvement per vessel through performance routing powered by AI.

But we know that, together, we can increase these numbers – and that collaborating to share approaches to handling distributional shift can not only make shipping greener, but also help sufferers of Multiple Sclerosis, help accelerate the autonomous car revolution, and help improve the lives of millions.

Launching The Shifts Project

To that end, we have launched The Shifts Project alongside a group of researchers from the Universities of Basel, Cambridge, HES-SO Valais and Lausanne. This is an international collaboration dedicated to studying and improving approaches to the critical topic of distributional shift in high-stakes, real-world settings (such as shipping).

Antonis Nikitakis, our AI Research Director, told me “This exciting collaboration will help increase the real-world impact of data-driven approaches such as DeepSea’s. Importantly, it will also provide a benchmark for shipping companies to understand where the industry is in terms of AI – and the sort of progress they should be expecting from their collaborators.”

Test your (virtual) metal

On that note, the Shifts Project has also launched the Shifts Challenge, an open initiative which encourages AI researchers – industrial and academic – in the shipping domain to test their own models’ capacity to handle real-world shipping challenges. The winners and runners-up will receive awards and – above all – a serious “stamp of approval” from some of the world’s top experts in the field.

More information on the Shifts Challenge can be found on the Shifts website:

What is distributional shift, and why does it matter in shipping?

Have you heard of distributional shift? Are you involved in shipping? If the answers are (no) and (yes), you need to read this!

Unless you happen to be one of the small (but increasing) number of data scientists in the shipping domain, chances are you haven’t heard of this phenomenon. But, in fact, it’s one of the greatest obstacles to using data really effectively – not only in maritime, but across every industry.

As our co-founder Konstantinos says,

Robustness to distributional shift is the difference between AI that works in theory, and AI that works for customers.

Distributional shift is a pretty big term for a pretty simple thing. And in shipping, it’s critical.

Let’s think about hull fouling. We all know what it is – but consider what this marine growth does to the data that is collected from your ship. The level of fouling affects some vitally-important things, including (1) the resistance of the hull; (2) the power required by the engines; (3) the impact of currents; (4) the fuel consumption; (5) the level of emissions.

To make matters worse, all these things – alongside many other data points – are interrelated in a very complex way. We can think of this as the “DNA” of the ship. With me so far?

To make matters EVEN WORSE – this entire complex web of data is changing over time, as fouling develops, in a largely non-predictable way. We can all agree – this makes understanding the data difficult. 

The same data phenomenon is seen in data sets from diverse sectors throughout the world.

That’s distributional shift! … so what?

Understanding the data shows us how to significantly increase vessel efficiency – through voyage and vessel optimisation – and the best tool in the world to make sense of this data is artificial intelligence. 

However, how a solution provider handles distributional shift can makes the difference between an approach that provides a measurable improvement to your vessels’ fuel consumption, and one that could even have a negative impact..

At DeepSea our pioneering AI research in distributional shift over the past five years powers our Pythia and Cassandra platforms – enabling, today, roughly an 8% average efficiency improvement per vessel.

Every day we have the great pleasure of working with shipping companies that are passionate about adopting new technologies to make their organisations greener, leaner and better connected. However, we also see that there is still a lack of information available to help aspiring companies make decisions about what to adopt, why, and what to expect.

Our technology manifesto

That’s why DeepSea follows a three-step ‘technology manifesto’ with you in mind:

  1. Phase 1: Pioneering research into AI approaches in shipping
  2. Phase 2: Proving the real-world utility of these approaches
  3. Phase 3: Advocating for common and transparent standards throughout the industry 

So now you know why distributional shift is so important! If you want to dig deeper – have a look at our exciting project alongside world-class researchers from the Universities of Basel, Cambridge, HES-SO Valais and Lausanne – The Shifts Project.

For more information on proving the real-world utility of AI approaches, check out our work on model validation.

Moving towards sustainable shipping: Key takeaways from Posidonia 2022

More than 27,000 people from the global shipping industry gathered last week in Athens, Greece for a full week of engaging discussions regarding the major challenges that the shipping industry is facing. Much of the discussion was focused on decarbonisation and the role that emerging digital technologies play in moving toward a net-zero future – key topics of the flagship panel sessions held by Capital Link and TradeWinds, where DeepSea participated.  

Here are some key takeaways:

Data explosion and environmental regulations necessitate the need for digital transformation.

Millions of data points can be generated from vessels. Huge databases are needed to manage the vast amounts of data, and skilled people are required to decipher it and turn it into meaningful and actionable insights. The current wave of emerging technologies is a valuable asset to the industry throughout this process. Close collaboration and communication between shipping organisations and start-ups can lead to the development of solutions that meet the pressing needs of the shipping industry.

In the past, the industry’s focus was on efficiency and cost minimisation / profit maximisation. Today, the third key pillar is emission reductions, and the ability to minimise disruption to the fleet when the new environmental regulations kick-in next year. Digitalisation is an important enabler of these objectives. What was done in the past is today done better and more efficiently using digital technologies. When shipping companies know what they want to achieve with strategic technology partners (eg. improving CII ratings and/or reducing fuel consumption), it makes it much easier to find a way to assess, monitor and measure the impact the technology is making.

DeepSea’s CEO, Roberto Coustas, speaking at the Capital Link Forum at Posidonia 2022.

Technology is no longer a luxury. It is an immediate and mandatory need for shipping organisations.

Applying new technology and changing an existing mindset in a traditionally conservative industry such as shipping can be a challenge. Facing this challenge, some shipping organisations which embrace innovation have created pilot technology programs involving their vessels and fleets – a critical element when developing solutions that address shipping’s most important needs. The shipping industry is slowly starting to realise the benefits achieved through technology, such as reducing CO2 emissions – and simultaneously saving on bunker fuel and optimising cash flow for their organisations. With the new environmental regulations in mind, harnessing proven digital technologies is no longer a matter of choice – but of urgent need.

Digital technologies also help bridge the gap between the crew aboard a vessel and the people ashore. With new and emerging technologies, machines are increasingly helping people to make better decisions – not based on intuition, but based on insights from data. Nevertheless, training people adequately is an increasingly vital part of adapting successfully to the digital era.

DeepSea’s CEO, Roberto Coustas, speaking at the TradeWinds Shipowners Forum at Posidonia 2022.

Validate this! We’ve published new research to help understand and judge the real-world utility of AI in shipping

At DeepSea we have a no-bullshit approach to AI. Today we are pleased to publish a new piece of research outlining a pioneering way of verifying the accuracy – and therefore utility – of a ship’s AI-generated model in real-world conditions. This is important – the more accurate the virtual model, the more efficient a ship can be made, and vice-versa.

Artificial Intelligence. It can do many things: understand whalesong, compose film scripts on-demand, and decrease the fuel consumption of a ship at sea. However, “AI” has also quickly become the greatest buzzword of the 21st century – wooing customers, investors and governments alike. 

It’s vital that serious researchers working to popularise this exciting approach continue to pursue rigorous methods of proving the real value of what they’re creating. In fact, we believe every end-user looking to employ this sort of service, in shipping or any industry, should demand it.

No-bullshit AI

At DeepSea we have a no-bullshit approach to AI. Today we are pleased to publish a new piece of research outlining a pioneering way of verifying the accuracy – and therefore utility – of a ship’s AI-generated model in real-world conditions. This is important – the more accurate the virtual model, the more efficient a ship can be made, and vice-versa.

The new approach was developed by seven of DeepSea’s thirteen-strong team of research scientists headed up by Dr. Antonis Nikitakis, and presented at the 2022 HullPic Conference in Tullamore, Ireland. 

The few models that currently provide an estimation of their accuracy all do so based on testing with data obtained from the same distribution (i.e. representative of similar conditions and containing similar biases) as the data used to train the model. For example, if the model is trained on data from the vessel’s historical behaviour, in a narrow range of well-experienced wind speeds or drafts, it is also tested on data with these speeds and drafts. Thus, the tests performed can’t tell if the model is reproducing the biases in the training data – and whether it will work as well in different, never-seen-before conditions. As anyone familiar with maritime data will know, real ship-at-sea data is actually highly variable. Most model accuracy figures reported in publications and marketing materials thus bear no relation to the actual utility of those models in real use cases.

DeepSea has long researched approaches to solving the technical challenge of boosting models’ ability to understand unseen (“out-of-domain”) conditions. However, before today, there has been no benchmark for evaluating this sort of competence within a vessel model. With this announcement, we are signalling that this rigorous test is a key part of our AI methodology. Moreover, we are releasing the details of the approach for global researchers to utilise themselves, in the hope of catalysing greater transparency across the industry.

“This research is an important step in helping our customers and the wider market to understand the true power, while alleviating the limitations, of an AI-based approach“, said Dr Nikitakis. “Coupled with the daily real-world impact we’re seeing on fuel consumption and CII ratings, we believe this sort of information is key to popularising this incredible technology throughout the industry.

A three-step approach with you in mind

This sort of research constitutes the second phase of DeepSea’s three-step approach to Artificial Intelligence in shipping:

  1. Phase 1: Pioneering research into AI approaches in shipping
  2. Phase 2: Proving the real-world utility of these approaches
  3. Phase 3: Advocating for common and transparent standards throughout the industry 


Having, and promoting, such a framework is especially important in the Shipping space, where early adopters are embracing AI for the first time.

Dr. Konstantinos Kyriakopoulos, CTO and co-founder of DeepSea, states: “We designed our AI framework for the direct benefit of the consumer. Once again, with this research I am so pleased we can fight the hype and support our no-bullshit approach to AI with such compelling evidence. It is exactly what DeepSea was founded to do, and every day it makes an increasing impact on our clients’ bottom-lines, and the sustainable future of the planet.”

Read the full paper

The full paper can be read here: On the evaluation of uncertainty of AI models for ship powering and its effect on power estimates for non-ideal conditions.

We’d love to discuss how the DeepSea approach can form part of your company’s sustainability strategy – we look forward to you getting in touch.