Sqwish raises $3 million for a decision model customers can keep training

Nov 18, 2026
The UK start-up's kettei fat model comes with post-training tools that use a customer's own data.

Sqwish, a UK AI start-up behind a decision model called kettei fat, has raised a $3 million pre-seed round, TechCrunch reports.

Co-founders Ushnish Sengupta and Federica Freddi met as graduate students at Cambridge and had been building small language models before realising their architectures could be repurposed as a decision model. Like other builders in the category, Sqwish has trained with synthetic data, but its main pitch is built-in post-training tools that let customers keep improving the model's judgment with their own production data.

"What makes us not a commodity, I think, is the fact that we are offering our customers the same power," chief executive Sengupta said. "They can keep uploading their production data, and just the same way that we trained our models, their models can also keep getting better at the tasks that they care about."

Kettei fat has performed well on some benchmarks, but Sengupta said benchmarks are "more of an invitation to try the model out, and then the continual tuning is the reason why they should stick to us," according to TechCrunch, which reported the round as part of a look at growing use of decision models.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like AI Funding Brief, AI in Europe or Daily AI Brief.

Also, consider following us on social media:

Free newsletter

AI in Europe

Daily AI news from Europe, chosen by Rene van Pelt: European companies, the EU and the UK, chips and data centres, deals and regulation, in a few minutes.

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more
Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.