Channel Capital at the Structured Finance Hackathon 2026

Kris Wilson, Chief Operating Technology Officer | Channel Capital

It was a privilege to attend the Structured Finance Hackathon 2026 in Barcelona, and an even greater one to help make it happen.

Channel Capital was proud to sponsor this year’s event, bringing together over 75 developers, data scientists, and AI specialists to tackle real challenges facing our industry. Organised by AI Tinkerers and timed to run alongside Global ABS in Barcelona, the hackathon gave participants just two weeks to go from idea to working prototype. 

This was the third edition, and the ambition of the field continues to grow. We were delighted to see engineers, quantitative analysts, and data architects from across the industry, including institutions such as Barclays and JPMorgan, converging on Barcelona to build, compete, and exchange ideas. 

Clive Humby coined the phrase “Data is the new oil”. In structured finance, we think that is exactly right, and behavioural payment data is the highest grade crude there is.

“Credit scores are a useful starting point, but they are a snapshot, a summary of history compressed into a single number. The real proof is in the payments themselves: whether a borrower pays on time, how they behave under pressure, and when their behaviour starts to shift.”

Our challenge: world model for credit

As a sponsor, Channel Capital set one of this year’s three challenges, and it reflects a question we think about every day. 

At Channel, we believe that past payment behaviour is the single best predictor of future payment behaviour. Credit scores are a useful starting point, but they are a snapshot, a summary of history compressed into a single number. The real proof is in the payments themselves: whether a borrower pays on time, how they behave under pressure, and when their behaviour starts to shift. Movements away from a borrower’s own typical pattern are often the earliest signal of stress or change, long before a credit score reflects it. 

Data tapes capture exactly this: rich behavioural sequences across loan payments and invoice settlement, covering arrears, cures, defaults, collateral updates, recoveries, prepayments, early payment frequency, and creeping delays, played out over time across thousands of borrowers. Yet most credit models treat these sequences as static inputs rather than the dynamic, context-laden signals they actually are. 

We asked participants: how might we build an open, GPU-accelerated foundation model that learns the full context of credit from these sequences, so that the resulting embeddings can be reused across default prediction, anomaly detection, and cash-flow forecasting? 

Structured finance sits on some of the richest longitudinal behavioural data in finance, yet the tooling to exploit it at scale remains fragmented and proprietary. Transferable, open credit representations would lower the barrier for every firm in the sector. We wanted to put that question in front of some of the best builders working in this space today, and the standard of entries did not disappoint.

Kris Wilson on the judging panel

Kris Wilson, Channel Capital’s Chief Operating and Technology Officer, served as one of three judges on Demo Day, evaluating finalist prototypes for technical rigour, practical applicability, and the quality of their structured finance domain reasoning. Congratulations to the winning team and to all the finalists, and a particular thank you to Luca Borella from Algorithmica for bringing Channel Capital into the event. 

What excites Channel Capital about events like this

We believe AI is going to transform private credit. The volume and complexity of data flowing through structured finance markets has long outpaced what is humanly possible to interpret at scale, and that is where AI has the most to offer. From identifying early behavioural signals in loan tapes to stress-testing portfolio assumptions across thousands of scenarios, the opportunity is significant and still largely untapped. 

Hackathons are one of the best formats for accelerating that progress. They bring together people with different skills and perspectives, and force them to deliver working proof of concepts against real-world challenges, under real constraints. The quality of thinking on show in Barcelona was striking, particularly around how to handle hallucination in LLMs when the outputs carry genuine financial consequences. The most interesting approaches combined deterministic methods alongside LLM-powered reasoning, using rules-based logic and verifiable computation to anchor outputs, with LLMs applied where interpretation and pattern recognition add value. That hybrid approach reflects exactly the kind of applied, domain-aware thinking that moves the industry forward. 

We were genuinely inspired by what the teams achieved. The creativity, technical depth, and domain understanding on show in just two weeks was a reminder of what becomes possible when the right people are given the right problem. Thank you to everyone who joined us in Barcelona, and to the AI Tinkerers team for another outstanding edition. 

If you were at the Hackathon in Barcelona (or missed it and want to find out more), or if you are working through similar questions at your own firm, I would be glad to connect.