Vetted Retrieval-Augmented Generation (RAG) Professionals

“Built and shipped a production "Campaign AI" multi-agent system (LangGraph) that personalizes B2B outbound emails at scale using Apollo.io prospect data, clustering-based segmentation, and 21 persona variants. Notably uncovered that high click rates were largely email security scanners and created a validated bot-detection/scoring pipeline (timestamps/IP/user-agent/click patterns), bringing reported engagement down from ~40% to a trusted 5–8% that aligned with real conversions.”

“Robotics software engineer who built an end-to-end system ("justmatrix"), focusing on multi-agent orchestration and a multi-RAG retrieval backend/API. Has hands-on ROS experience, including a custom node for reliable high-frequency sensor data routing, plus deployment automation using Docker, Kubernetes, and CI/CD.”

“Data/AI practitioner who built a production LLM-driven healthcare claims analytics and dashboarding system to reduce avoidable ER visits—processing 1.4M+ claims, flagging 19% as non-emergent, and projecting ~$2.8M in annual savings. Demonstrates strong real-world LLM reliability and performance engineering (grounding, numeric validation, caching, materialized views, quantization) plus orchestration experience with Airflow and Azure Data Factory.”

“Built and scaled an internal AI code-search/assistant agent that expanded from engineering-only to broader internal users, tackling legacy code and inconsistent standards to make a RAG pipeline production-ready. Uses a metrics-driven approach (user feedback + automated Python evaluation for retrieval relevance and latency) and has handled high-pressure outages, including moving parts of the stack off AWS and adopting Milvus on internal infrastructure for resilience.”

“AI/backend engineer who has built and scaled production LLM-powered SaaS features (document assistant + compliance review agent) on a Node.js/TypeScript + Postgres/Redis stack deployed to GCP Kubernetes. Demonstrates strong production reliability chops—async queueing, autoscaling, observability, and database tuning—with quantified wins (p95 latency -60%, query 4s to <200ms) and robust AI guardrails (strict RAG, schema validation, citations, HITL).”