By Sonali Jha | Founder & CEO, Cunomial Technologies
The maritime industry isn’t short on data. What it’s missing is a trusted, repeatable way to turn that data into something useful.
Every vessel throws off operational data as it runs. Ports track cargo flows, logistics, infrastructure signals. Classification societies sit on decades of engineering knowledge. Universities and startups keep building smarter AI tools. But the companies holding the valuable data usually aren’t the ones building the next generation of technology.
Ship owners don’t want to hand over commercially sensitive information. Innovators need real evidence their tools actually work. Port operators and engineers want proof that a new system won’t quietly undermine safety or control. That’s usually where good ideas die.
More algorithms won’t fix this. Neither will more hackathons. What’s needed is a coordinated approach: pair a controlled maritime AI data sandbox with real innovation management. One gives you a secure place to test things. The other gives you a disciplined way to push the best ideas toward actual deployment.
This came up a lot at the Maritime AI Conclave 2026, hosted by the Indian Register of Shipping (IRS) and the Centre of Excellence in Maritime and Shipbuilding (CEMS) in Mumbai. I dug into it further in a co-authored piece for the event magazine, Maritime AI Horizon 2026, titled “Sharing Without Surrendering: A Framework for the Maritime AI Data Sandbox.”
An ocean of data is of little value if innovation cannot access it
Maritime AI has real potential here: predictive maintenance, vessel-performance analysis, inspection prioritization, port coordination, document-heavy workflows. But a model is only as good as the data behind it. Fragmented datasets, inconsistent formats, and locked-down access can kill a project before anyone even gets to test it.
Start with a practical question: what decision needs to get better, who actually owns that decision, and what evidence would make them change course? “Use AI in shipping” is just a theme. “Help engineers flag which structural spots need a closer look, based on historical thickness readings” is a problem you can actually test.
Sharing without surrendering: the maritime AI data sandbox
A maritime AI data sandbox is a controlled space where approved teams can build or test solutions against real maritime problems, under rules everyone agrees to upfront on access, security, and what comes out the other end. Early-stage teams might only get synthetic data and a data dictionary. Shortlisted teams might get restricted historical data, but still inside a locked-down environment. The strongest solutions can graduate to a time-bound pilot.
Sometimes the computation itself happens inside an environment the data owner controls, and only the approved output ever leaves. That matters because stripping out vessel names doesn’t actually make data anonymous. Routes, dates, technical quirks, and unusual incidents can still give things away. Synthetic data isn’t a free pass either. It needs its own fidelity and disclosure-risk checks.
Share the intelligence. Not the vulnerability.
Access should be staged, never assumed. Every step up needs clear criteria, someone accountable, and a paper trail of what got approved. A sandbox lets you experiment. It doesn’t get you regulatory approval, class certification, or a green light to deploy.
From AI experiments to enterprise adoption
A promising prototype isn’t the same thing as a trusted operational tool. Maritime companies need to measure new approaches against a documented baseline, protected test data, and criteria that actually make sense for the domain. Accuracy numbers only tell part of the story. False alarms, missed warning signs, how much integration work it takes, how much a human still has to double-check. All of that matters just as much.
Take a corrosion-risk example. Vessel age, structural location, thickness readings, coating condition. Feed that into a model and you can help engineers prioritise where to look closer. But the engineers still design the evaluation and judge the results. The people with the inspection and class authority are still the ones responsible for the final call.
A real pilot plan spells out who owns what, what the cybersecurity and integration requirements are, how success gets measured, how often you report back, who owns the IP, how long data gets retained, and under what conditions you keep going, change course, or pull the plug. That’s what turns a cool demo into a decision someone can actually defend.
Why maritime AI needs innovation management, not just a sandbox
The sandbox solves the data problem. It doesn’t solve the people problem. Someone still has to publish the challenge, review applications, assign evaluators, decide who moves forward, track milestones, and figure out whether a pilot actually delivered anything.
That’s where a proper innovation management platform earns its keep. Programme management, challenge management, expert evaluation, mentoring, milestone tracking, impact reporting. All of it needs to run as one connected system, not a pile of spreadsheets and email threads nobody can find later.
The sandbox makes experimentation possible. Innovation management is what actually moves things forward. Put them together and you’ve got a real path to adoption.
A working pipeline usually looks something like this: define the industry challenge, open up controlled data access, run technical validation, bring in expert evaluation, run an operational pilot, then make the deployment call. Each stage should leave behind evidence, not just activity.
Where Accubate fits: enterprise innovation management at scale
At Cunomial, we build Accubate, an enterprise innovation platform institutions use to run complex innovation programs. It handles applications, evaluations, cohort management, mentoring, grants, milestones, events, and impact reporting, all through configurable workflows.
For a maritime AI challenge, Accubate can handle the programme side: onboarding participants, running expert reviews, managing cohorts, scheduling mentors, tracking pilot milestones, keeping decision records straight. The actual protected data environment is a separate job. Data custody, model execution, cybersecurity, export approvals, that needs its own specialist infrastructure and governance. Connecting the two systems would take a scoped-out integration project of its own.
We’ve done this kind of work before, including for FITT at IIT Delhi’s NCIIPC-Startup India AI Grand Challenge. FITT’s completion certificate credits Accubate for registration, applications, submissions, evaluator and jury workflows, the leaderboard, communications, scheduling, reporting, and centralised programme operations. That shows what the programme-management layer of a multi-stakeholder AI challenge looks like.
Across more than 200 institutional engagements, we keep seeing the same thing: innovation scales when participation, accountability, and evidence are actually coordinated. For maritime organisations, that kind of ecosystem management is often what’s missing between a promising idea and a decision backed by real support.
Building a global maritime innovation ecosystem
Maritime ecosystems around the world can pull shipowners, ports, classification societies, research institutions, startups, accelerators, and maritime global capability centres into one connected innovation model. Shipowners and ports bring real operating problems and pilot settings to the table. Engineering specialists define what safety and evidence actually need to look like. Universities bring research depth. Startups bring new ideas worth testing.
GCCs serving maritime companies are well placed to anchor this. Not as technology delivery shops, but as strategic innovation hubs in their own right. Whether they’re based in an established shipping hub or somewhere newer in tech, they can connect global shipping companies, port operators, classification societies, research institutions, and startups. Combine domain expertise, AI engineering, and real innovation management, and GCCs can run industry-led challenges, support collaborative research, back startup acceleration, and help commercialise new technology. Accubate can support the programme workflows, evaluations, milestones, and reporting that hold all of that together.
A first programme doesn’t need to be huge. One committed sponsor, one data custodian, one well-defined use case, and a real budget is enough to get started. Track more than just registrations. Look at how long review and data-access approvals actually take, how far pilots progress, and how outcomes compare to baseline. The goal isn’t more demos. It’s faster, better decisions.
The maritime sector won’t be judged by how many pilots it launches. It’ll be judged by how many of those pilots turn into something people actually use.
The next breakthrough is institutional
Maritime AI moves forward when industry expertise, real data access, technical evaluation, and accountable programme management all work together. A well-built data sandbox creates room for responsible experimentation. An innovation management platform organises the challenge, the collaboration, and the progression around it.
The future of maritime AI isn’t just about smarter systems. It’s about building ecosystems that can keep turning hard problems into solutions that actually work.
If you’re planning a maritime AI challenge, a research collaboration, an accelerator, or an innovation programme, start with the actual operational problem you’re trying to solve, and figure out what evidence you’ll need to act on it.
Explore Accubate: accubate.app | Cunomial Technologies: cunomial.com
About the author
Sonali Jha is the founder and CEO of Cunomial Technologies. She holds an M.S. in Computer Science from NYU’s Courant Institute and has worked in financial technology and computational genomics.
References and further reading
- Jha, Sonali. “Sharing Without Surrendering: A Framework for the Maritime AI Data Sandbox.” Maritime AI Horizon 2026.
- Lloyd’s Register and OneOcean, Mastering Maritime Data for Competitive Advantage.
- BIMCO, Shared Principles for Maritime Digitalisation.
- DNV, Industrial AI Regulatory Sandbox Consortium.
- UK Information Commissioner’s Office, Introduction to Anonymisation.