Work — Real results from real builds

We ship useful systems and share the receipts. Here are a couple of highlights.

What these builds have in common

Fast delivery

Weeks, not months, from concept to production

Measurable outcomes

Clear before/after metrics you can track

Real business reliability

Built to handle real customers, not just demos

Open Source

Beyond client work, we contribute to the AI community.

AI Research

ToolGRPO

Teaching LLMs to use tools through reinforcement learning. A modified GRPO implementation that trains models to solve complex problems by writing and executing Python code.

Reinforcement LearningLLM Fine-tuningTool Use
View on GitHub

Want something like this?

Let's plan a first project that delivers real value quickly.