Aanya Raj Singh

At Dürr Group, I co-architect a multi-model AI chat platform handling 70K requests per month and 20M tokens per day across five models. I build production LLM features including RAG/Q&A, structured outputs, agentic web search, summarization, comparison, and bulk CSV workflows.

I built an eval-driven RAG pipeline using Azure Document Intelligence, HNSW hybrid search, and metadata filtering, improving retrieval accuracy by 60% and latency by 50% in offline evaluation. I also developed a Salesforce service-case ML system for 95K+ tickets, improving routing accuracy to 71.0% and enabling confidence-calibrated auto-routing projected to save 1,700 hours annually.

Earlier, I improved PySpark ingestion speed by 20%, conducted EDA across 200+ GB of data, and delivered 20+ freelance AI and ML projects across healthcare, finance, and e-commerce. I take systems from data engineering and prototypes through reliable production deployment.