Biography

An AI engineer who takes LLM systems from prototype to production. He built an Indonesian legal AI platform used by thousands of people, with hybrid retrieval that beats the baseline in 27 of 40 blind evaluations.

He also researches LLM fine-tuning for Indonesian NER that beats the baseline by 7.17 F1 points, and served as President of BEM Fasilkom UI.

Achievements

  • President of BEM, Faculty of Computer Science, Universitas Indonesia, after serving as Head of the Student Advocacy and Welfare Department
  • Teaching assistant for Operating Systems and Discrete Mathematics 1, Universitas Indonesia

Projects and experience

Klaussa

AI Engineer

Indonesian law is spread across dozens of government sites and hard to trace. Klaussa brings it together in one AI legal research platform with citations straight to the article and source document, so every answer can be verified. Evans built this platform (FastAPI and React) on the AI, retrieval and production reliability side: a multi-agent research assistant with a planner-supervisor-executor loop, 15 tools and a 7-phase pipeline; hybrid legal retrieval over 300k+ regulations (BM25, dense vector, HyDE, RRF fusion, cross-encoder reranking) that wins 27 of 40 blind evaluations; production hardening (JWT validation, login lockout, multi-key Gemini rotation with failover); and release safety with 2,200+ automated tests, OpenAPI contract-drift checks, Trivy vulnerability scans and auto-rollback. The platform indexes 3.1 million+ articles and 319 thousand+ legal documents, has answered 6,900+ legal questions, analyzed 1,250+ documents and saved 8,000+ hours of user work, and is used by legal teams, advocates, students and B2B clients.