Chaitanya Sachdeva

About

LLM engineer who has deployed privacy-preserving, real-time workplace risk monitoring over massive enterprise chat/email streams, tackling latency, hallucinations, and extreme class imbalance with model benchmarking, RAG + fine-tuning, and a pre-filter alerting layer. Also built an agentic legal contract drafting system (Jurisagent) using LangGraph/LangChain with deterministic multi-agent control flow, structured outputs, and reliability-focused evaluation/telemetry.

Experience

Education

Key Strengths

• Built and deployed real-time LLM-based workplace risk monitoring for enterprise chats/emails
• Model selection under strict latency constraints (tested 10+ open-source models; chose Llama 3 7B)
• Reduced false positives and improved real-time performance via a 0/1 pre-filtering alert mechanism
• Mitigated hallucinations/false positives using RAG plus fine-tuning when retrieval lacked relevant context
• Designed privacy-preserving architecture using on-device/on-prem open-source LLMs to keep data on company servers
• Built agentic, multi-step contract drafting system with planning/drafting/reflection agents using LangGraph + LangChain
• Reliability-focused evaluation approach: fixed eval sets, schema validation, regression tests, and production telemetry/KPIs
• Effective collaboration with non-technical HR/legal stakeholders by translating ML concepts into risk levels, alerts, and dashboards