What we build

  • Knowledge and retrieval systems

    Search, question answering and agents over your documents and databases, built on retrieval and knowledge graphs.

  • Private deployment

    Open models served inside your own network with vLLM or Ollama, and tuned to the hardware you have. For data that can't go to a third-party API.

  • Architecture review and evaluation

    A written review of your design, model choices and running costs. We test candidate models against your own tasks before you commit to one.

Your networkYour dataOpen modelvLLM or OllamaYour hardwarethe model is tuned to itThird-party APInot used
Your networkYour dataOpen modelvLLM or OllamaYour hardwarethe model is tuned to itThird-party APInot used

Architecture review

Datawhere it lives todayDesignsystems and constraintsModel choicestested on your own tasksRunning costsknown before you commitYou receiveA written review
Datawhere it lives todayDesignsystems and constraintsModel choicestested on your own tasksRunning costsknown before you commitYou receiveA written review

For investors

The same review, done before you invest. We read a company's design, data and model choices, test its claims against its own tasks, and write down what holds and what doesn't.

How an engagement runs

Tell us what you're building

Contact us

See our work ›