01 · Retrieval
Private knowledge
Small, grounded search layers over internal documents — with citations, not confabulation.
Applied AI · since 2024
Small models.
Quiet infrastructure.
We build focused AI tools for teams that prefer calm systems over noisy demos: retrieval, evaluation and automation that stay close to the data.
02 · Evaluation
Quiet benchmarks
Task-specific eval harnesses that tell you when a model is actually better, and when it is not.
03 · Automation
Boring agents
Deterministic workflows around the model: retries, guardrails, logs, and a human in the loop.
Current focus. A compact retrieval stack for engineering archives, running on a single node in Frankfurt. No dashboards for the sake of dashboards — just fast answers with a traceable path back to the source.
lat 50.11 · lon 8.68