Skip to main content

Sqwish

Sqwish sells decision models, called kettei, that pick among answers you give them instead of writing text. Each model returns a probability for every option, so a support agent or contract review tool can accept, escalate or reject a call. The models come in four sizes and can be fine-tuned on a company's own past decisions.

Traction 10/10
Cambridge, United Kingdom · HQ
Founded 202482K+ followers
  • AI Agents
  • Artificial Intelligence
Updated today

Funding

Raised to date

$5.9MRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.

across 2 rounds

CEDIULG+3

$3M · Pre-Seed

Announced Nov 2026 · from article date

Disclosed cumulative: $5.9M

2/2

Founders

2 founders

Ushnish Sengupta

Co-founder & CEO

Cambridge, United Kingdom

  • Cambridge

PhD in Engineering at University of Cambridge

Federica Freddi

Co-founder

Cambridge, United Kingdom

  • Cambridge

MEng in Engineering at University of Cambridge

Product

Problem

AI agents use large language models for steps that only need a choice among options the agent already holds, such as picking a tool or judging an answer. The model writes out an answer in text, which costs tokens and latency, and returns one sample rather than a probability for each option.

Solution

Sqwish makes decision models: you give the model a context (text, images or both), a question and the allowed answers, and it returns a probability for each answer in a single pass. The kettei family comes in four sizes, from kettei fat for the hardest calls down to kettei dot, which runs on your own device. The two larger models read images as well as text. Customers fine-tune a model on decisions they have already made, and the company says a tuned model can be ready in about 15 minutes. In production, each outcome feeds the next round of training, and a new model joins the family only if it passes its gate.

Target Audience

Teams building AI agents and AI products that need fast, cheap choices among fixed options, such as support automation, contract review and safety checks.

Features

  • Get a probability for every allowed answer, so you can set a threshold and decide whether to accept or escalate.
  • Fine-tune on your own past decisions to teach the model your contracts, audience or rules.
  • Run the smallest model that gets the decision right, from kettei dot on-device up to kettei fat.
  • Use the same model for different choices, such as triaging a ticket, ranking passages or picking a tool, by describing the decision in plain language.
  • Run kettei core and kettei dot yourself, since both are open source.
  • Keep improving the model in production as new outcomes arrive.
This profile is AI-generated and may contain inaccuracies.