Datalaked Ltd. · Sofia / Gorna Oryahovitsa, Bulgaria
We design and deliver custom SaaS solutions with the depth of a dedicated team and the agility of a studio. Alongside client work, our AI Labs division runs open research on real production systems.
What we do
End-to-end product development for businesses that need software shaped around their process — not the other way around. FastAPI, React, PostgreSQL, and whatever the problem actually calls for.
Small team, direct communication, no handoff to juniors. You work with the people who understand your domain. We stay close through design, build, and after go-live.
Datalaked AI Labs
Most AI benchmarks use synthetic tasks. We run evaluations on production codebases — real migrations, real bugs, real sprint backlogs. The results are published openly so the community can replicate and extend them.
An open methodology for benchmarking LLM coding agents on real production code. 13 task definitions across 4 categories, a 0–5 scoring rubric, and setup guides for any model via Claude Code + LiteLLM. Run it on your own codebase.
Open Source MIT LicenseContact