Chaitanya B.

I build production LLM systems for legal and fintech clients, from RAG pipelines to OCR extraction to full extraction pipelines. Systems that run in production and hold up under real data. Some of what I've shipped: Built a legal RAG chatbot over SEBI and FIU regulatory documents. Handled the full stack: OCR chunking across four different engines, hybrid BM25 + dense retrieval, citation extraction for legal provisions, and security hardening. Migrated embedding models and re-indexed a corpus without breaking retrieval quality along the way. Built a production invoice extraction tool for a New Zealand client. Multi-tenant SaaS on AWS, GPT-4o extraction, business rules registry, human in the loop verification UI so low confidence fields get reviewed before anything is finalized. Targeted 95%+ accuracy on live invoices and debugged the messy real-world stuff: watermark contamination, vendor field swaps, tax ID misattribution. Outside client work, I'm constantly building side projects because I like getting hands on with new tools before they're mainstream. That habit is where a lot of my practical RAG and OCR knowledge actually comes from. Before freelancing, I spent two years as a backend developer at Accenture on client-facing engagements (Spring Boot, Angular), promoted from associate to analyst in 15 months. I hold an MSc in Data Science & Analytics (Distinction) from Cardiff University. I'm currently working full-time with a consulting team leading LLM application and agentic workflow builds for different clients with different problems. If you need someone who can take an ambiguous "we want AI to do X" and turn it into something that actually runs, works, and holds up in production, that's what I do.