Athang Bachhav

I’m an ML Engineer with 4+ years of building and shipping production ML systems—turning messy data and real business needs into reliable pipelines, decision support, and user-facing intelligence. I specialize in LLM/NLP applications, feature pipelines, and low-latency model serving.

In my recent work, I designed and deployed an automated LangChain pipeline to cut leadership reporting time from 3 days to 4 hours by replacing a manual Excel-and-email workflow. I also built a RAG-based knowledge portal (Milvus vector store + GPT-4) that indexed 5 years of policy and program documents, enabling 200+ staff to self-serve institutional knowledge and reducing research time from 45 minutes to under 3 minutes.

I focus on performance and operational correctness in production. I improved outreach targeting precision by 34% using spaCy and Transformer-based NLP classifiers trained on 12,000+ survey responses, and I reduced model endpoint error rates from 8% to 0.6% by adding structured request validation, latency logging, and alerting across multiple Flask + Docker microservices.

Previously at AIG, I built a LangChain + Pinecone RAG system that reduced underwriter review time by 60% by auto-extracting key contract clauses from 50,000+ insurance PDFs, and I trained XGBoost and Random Forest ensembles to lift risk classification precision from 71% to 89%. Across both roles, I partner closely with stakeholders to translate goals into scoped ML work with clear success metrics—and I care deeply about reducing manual effort while keeping latency low.