MIKE HOANG TRAN
Welcome to My Portfolio
Building Production AI Systems & LLM-Powered Applications
Senior AI Engineer
RAG Pipelines • Vector Search • LLM Applications • Python • Production AI Systems
About Me
Building Production AI Systems & LLM Applications
Applied AI engineer with 7+ years of Python/TypeScript experience, specializing in RAG systems, LLM-powered applications, and production AI infrastructure. I build intelligent assistants, document Q&A systems, and vector search pipelines that deliver accurate, grounded answers with proper guardrails and human-in-loop controls.
From prototyping with LangChain and GPT APIs to deploying production FastAPI services with Terraform and monitoring—I deliver full-stack AI solutions that integrate seamlessly with existing workflows. Skilled at building end-to-end systems that combine ML models, data pipelines, and backend services with a focus on reliability, scalability, and business impact.
Experience Timeline
My journey in building production AI systems and intelligent applications
Senior AI Engineer
CareFirst BlueCross BlueShield
Nov 2024 – Oct 2025
Built "ClaimGuard", a GPT-4o/LangGraph assistant handling policy Q&A for 8k internal users; answer times dropped 42% with citations and flag buttons. Wired LangGraph agents to read claims packets, route summaries through FastAPI tools, and ping human reviewers only when anomaly scores spike—saved ~500 analyst hours monthly. Put real guardrails in place: LangSmith tests, custom judge models, PII scrubbers, and dashboards; accuracy climbed 18% quarter over quarter. Merged Azure Cognitive Search, Pinecone, and 1.2 TB of docs into one nightly-refreshed RAG index; grounded answers stayed above 90%.
Applied AI Engineer
MobiDev
Jul 2022 – Jun 2024
Built a Python "automation hub" pairing ML forecasts with GPT drafts for exec updates; reports that took days now take hours with forecasting error around 2%. "Diagnostics Copilot": LangChain + Weaviate chatbot for model drift questions—support teams could ask "why did churn spike?" and get tagged plots plus next steps; investigation time fell 35%. Standardized prototyping: Airflow + MLflow + FastAPI pipelines to get bots/services into staging within 10 days, not multiple sprints. Crafted reusable prompt/eval harnesses (synthetic data, regression suites) shaving setup time ~40%.
Software Engineer
ADP
Nov 2019 – Mar 2022
Shipped React + GraphQL workforce dashboards with live anomaly alerts (embedding-based detectors in Python) for 25k users; false positives stayed under 3% and refresh was sub-second. Added chatbot-style widgets to payroll tools that surfaced "here's what changed and what to do" using internal APIs + lightweight NLP; satisfaction scores rose 18%. Built event-driven WebSocket services so HR teams saw payroll shifts in real time, processing 2k+ events/day without page reloads. Owned CI/CD and Terraform for supporting microservices, averaging 20+ releases each month with automated tests, feature flags, and CloudWatch alarms.
Tech Stack
Core technologies and frameworks for building production AI systems, RAG pipelines, and scalable backends
LLMs & Applied AI
Backend & Data
Frontend & Automation
Cloud & DevOps
Featured Projects
Production-ready AI systems featuring RAG architectures, LLM-powered applications, and intelligent automation.
InteliJuris - Legal Research RAG Platform
A cutting-edge legal research platform that transforms how legal professionals explore Mexico's SCJN jurisprudence. Built with advanced RAG architecture, it enhances the existing Buscador Jurídico by integrating Generative AI for intuitive, context-aware legal document search and retrieval.
Memomind - AI-Powered Note-Taking Assistant
A sophisticated note-taking application featuring chat-based RAG workflows and AI-powered insights. Built with modern web technologies, it leverages LangChain and Llama3 for intelligent content generation, with secure authentication and a polished UI built on Next.js and TypeScript.
Get in Touch
Open to opportunities in AI engineering, RAG systems, and production LLM applications