Cohabs AI Lab
An internal platform giving 100+ colleagues secure access to several AI models and to company data.
At Cohabs I founded and run the AI Lab: an internal platform that gives every team secure access to large language models and to the company’s own data.
The starting idea is simple. Without a shared foundation, every team that wants to do something with AI has to solve the same problems on its own: which model, how to reach it without exposing data, and how to connect it to internal systems. The AI Lab solves that layer once.
What the platform provides
- Multi-model access: people use the model that fits the job instead of one imposed tool
- Access to internal data through AI, without going through the engineering team for every question
- A place to experiment, where each department can test its own use cases
The infrastructure
It runs on a multi-cloud architecture, with Mastra on top of AWS, Azure and GCP, built for high availability and for keeping data hosted in Europe. RAG pipelines connect the models to internal documentation and operational data.
The hardest part was not technical. Shipping the platform was half the job; the other half was helping teams work out what to do with it.