Chandana J

I’m a Senior AI Engineer focused on GenAI and agent systems, building LLM-powered platforms that turn messy inputs into structured, dependable outputs. I specialize in RAG, agent orchestration, and reliability layers that make generated results auditable and production-ready.

At Courtroom Insight, I designed an LLM-powered Agentic Legal Analysis Pipeline that automates case briefings, analysis, and litigation preparation. Using LangChain and RAG, I orchestrated specialized agents via an IRAC workflow, transformed multi-level narratives into structured executive summaries, and improved legal narrative clarity and consistency.

I also built an agentic research assistant with RAG using OpenSearch vector search for discrepancy-focused legal Q&A. To reduce generation errors by ~60%, I implemented a reliability/validation layer with JSON schema and Pydantic, including retry logic and fallback prompting, plus RAG evaluation metrics (answer relevance, contextual recall, hallucination detection) with structured logging and monitoring.

Previously at Deloitte USI Consulting, I engineered a GenAI RFP Accelerator using AWS Bedrock (Claude 3 Sonnet) and OpenSearch to automate analysis of complex proposals. I implemented chunking/embedding strategies, embedding caching and prompt compression (reducing inference latency by ~25%), and a synthetic data pipeline with AWS Bedrock and Snowflake, alongside distributed ETL pipelines and validation frameworks for ML-ready investment data.