Raj Patel

About

Production-focused applied ML/LLM engineer who has deployed an LLM-powered RAG assistant and improved reliability through rigorous retrieval evaluation (recall/MRR), reranking, and guardrails that prevent confident wrong answers. Experienced running containerized ML/LLM services on Kubernetes (including AWS-managed layers) with CI/CD and observability, and has delivered a real-time predictive maintenance system using streaming sensor data and time-series anomaly detection in close partnership with maintenance teams.

Key Strengths

• Built and deployed production LLM RAG assistant with citations over company knowledge base
• Improved RAG answer faithfulness via chunking/metadata filtering, reranking, and evaluation sets
• Implemented guardrails to abstain or ask clarifying questions when retrieval confidence is low/contradictory
• Systematic component-level testing of retrieval vs generation with golden datasets and IR metrics (recall, MRR)
• Designed for reliability with failure-scenario testing (index outages, stale embeddings) and correct fallbacks
• Production monitoring focused on hallucinations, abstentions, latency, and user follow-ups to detect regressions
• Kubernetes-based orchestration of containerized ML/LLM inference services, APIs, and vector stores with CI/CD and observability
• Delivered predictive maintenance solution by aligning metrics/thresholds with maintenance stakeholders (downtime reduction, alert precision)

Experience

Education

Skills

Languages