Anuttara Rajasinghe

I've built production AI systems at Altrium, including semantic talent matching for an internal HR talent pool using RAG pipelines, embedding models, and candidate-to-role semantic search.

I own retrieval architecture across eight services, using hybrid dense-vector search, BM25, Reciprocal Rank Fusion, HyDE, and structured re-scoring. I also evaluated embedding models using precision@k and dimensionality trade-offs, and moved embedding cost to write time to eliminate query-time LLM token usage.

I've strengthened shared LLM platform reliability through structured-output parsing, retry logic, temperature-escalation policies, and provider failover across Gemini, OpenAI, and Azure AI Foundry. My work also includes Docker and Testcontainers-based vector-search testing, candidate engagement tracking, and resolving quality-engineering defects across talent matching and outreach.

Earlier at Altrium, I delivered LLM-powered RAG pipelines, multi-agent systems, voice-enabled GenAI bots, AI chatbot demos, and retrieval systems for insurance data. I'm pursuing an MSc in Artificial Intelligence Engineering while continuing to turn HR tech, insurance, and enterprise sales requirements into reliable AI solutions.