At Systems Limited, I designed and built generative AI solutions that ground LLM responses in authoritative documents, with inline citations and direct links to sources. I developed document ingestion pipelines and hybrid retrieval with reranking.
I implemented production guardrails for PII/PHI detection and redaction, scope restrictions, and refusal handling. I also designed evaluation frameworks covering retrieval quality, groundedness, hallucination rates, and calibration, and published related research in Applied Artificial Intelligence, Taylor & Francis.
At Personio, I built LLM-powered NLP features for document processing, semantic search, and structured outputs. My projects include a multi-provider LLM serving layer and an agentic document Q&A workflow; I’ve also worked with AWS deployment and GitHub Actions CI/CD.
