creit.
01Selected work

Some of the problems we've solved.

A selection of production AI systems, agent runtimes, and engineering platforms delivered for growing companies and enterprise teams.

AI Engineering

Making AI faster and cheaper to run

Problem
A growing product’s AI was getting slower and more expensive as more people used it.
What we did
Rebuilt how the AI was run behind the scenes, and trained a custom version on the company’s own work so it performed better for their specific needs.
Outcome
The custom-trained AI outperformed general-purpose tools for the company’s day-to-day use.
AI Engineering

AI that does the work, not just the chat

Problem
A financial services team spent too much time producing reports manually, and worried about mistakes.
What we did
Built an AI system that reads the relevant information, checks it against multiple sources, and drafts the report on its own.
Outcome
Built-in checks now catch errors and inconsistencies before a person ever sees the draft.
AI Engineering

AI that gives real answers, not guesses

Problem
A voice assistant needed to give accurate answers during live conversations, not make things up.
What we did
Connected the assistant to the company’s real, up-to-date information, so every answer it gives is grounded in fact rather than guessed.
Outcome
The assistant now gives accurate answers in real time, during live conversations.
AI Engineering

Making messy data usable for AI

Problem
A team had large amounts of raw data and images that were too disorganised to use in AI projects.
What we did
Built tools that automatically organise and prepare that data — including images and text together — so it’s ready for AI to use.
Outcome
Data that used to take weeks to prepare is now ready in a fraction of the time.
Software / Product

AI dev-tooling product

Problem
Engineering teams were losing time triaging and fixing routine issues.
What we did
Built an AI-powered tool that turns engineering issues into pull requests in minutes, integrated directly into the existing ticketing workflow.
Outcome
Cut manual ticket-fixing time for the engineering team.
Software / Product

Restaurant management platform

Problem
Restaurant operators needed one system instead of a patchwork of tools.
What we did
Built a full-stack web and mobile restaurant management platform covering staff, orders and operations.
Outcome
Positioned as the operator’s single system for running the business.
Software / Product

AI video search tool

Problem
Finding a specific moment in raw video footage meant scrubbing manually.
What we did
Built an AI video analyzer that takes a video and a plain-language description of the moment you’re looking for, with an optional reference image, and returns the match.
Outcome
Turned manual video review into a search query.
Software / Product

Consumer social app

Problem
A student community needed a way to make real friends, not just connections.
What we did
Built a mobile social app designed specifically for college students, from the ground up.
Outcome
Shipped as a product built by and for its own user base.
02// Capabilities

Tougher AI problems we solve

Making an AI agent reliably use internal tools and APIs without breaking things.
Building retrieval that finds the right document, not just a similar one.
Running open-source models efficiently instead of paying for every API call.
Automating a workflow that currently depends on human judgement.
Connecting AI systems to messy, inconsistent enterprise data.
Scaling inference across multiple GPUs without wasting capacity.
03// Engineering

Tougher software problems we solve

Modernising a monolith into services without breaking existing customers.
Making a real-time system — payments, matching, streaming — actually real-time.
Building for two audiences at once: end users and administrators.
Building a technical team fast enough to hit an aggressive roadmap.
Keeping a mobile app reliable across a long tail of devices and OS versions.