Minh Nguyen.

I build ML systems from zero to production. Fast.

Machine Learning Engineer with 3+ years at BlackRock, Tesla, and Homebase (Y Combinator). Strong foundation in taking ML systems from zero to production — spanning document intelligence, agentic reasoning, RAG pipelines, and real-time data infrastructure. AI-native workflow with Claude Code, Copilot CLI, MCP servers, and custom agent skills. I work best in high-ownership teams that value speed, reliability, and end-to-end execution.

Machine Learning· Agentic AI· RAG· Document Intelligence· MLOps
San Francisco· San Jose· Seattle· Redmond· New York· Philadelphia

Get in Touch

Cold emails welcome. Fastest reach is direct email.

Experience · 2Y FTE · 2Y+ Intern & Research

Built and shipped 4 production LLM and agentic systems end-to-end — automating 20,000+ hours of financial operations annually and serving 200+ internal users across RockAI.

  • Built FIESTA, an end-to-end document understanding pipeline that automatically processes and extracts structured data from invoices using LLMs, OCR, and embedding-based retrieval.

    LangChain Azure Document Intelligence Embeddings RAG Python Docker Kubernetes
  • Designed ValidAide, an agentic validation engine that reasons over documents and databases to automatically QC financial reports — replacing a previously manual process for a 30-person team.

    LLM-as-a-Judge Agentic AI LangGraph SQL Python Docker Kubernetes
  • Created AI Halo, an internal Python SDK that gives the org a standardized way to measure AI adoption, system performance, and business ROI across all RockAI products.

    Python Java Snowflake InfluxDB Grafana Power BI Docker Kubernetes
  • Shipped a LLM-powered ticketing platform that triages requests, generates actionable items, and automates leadership reporting for 200+ internal users.

    LLMs Streamlit Docker Kubernetes Azure DevOps

Shipped real-time ML infrastructure across vehicle manufacturing assembly lines — reducing downtime by 30% and giving engineers predictive visibility 3 days ahead of equipment failure.

  • Built a predictive maintenance system for factory robots — training ML models to forecast equipment failures before they cause production downtime, achieving 90% accuracy 3 days ahead.

    TensorFlow scikit-learn Time Series SHAP Python
  • Developed real-time anomaly detection systems across production lines to surface equipment failures and improve visibility for manufacturing engineers. Reduced downtime by 30%, optimized runtime by 500%.

    Go Python InfluxDB MQTT Grafana Docker
  • Trained regression models to simulate equipment behavior, giving engineers a data-driven way to refine and tune machinery across 200+ welding robots.

    scikit-learn MLflow Python CI/CD Jenkins

Designed and shipped two production ML systems from scratch at a YC-backed startup — a recommender engine on AWS and a risk automation platform with full MLOps infrastructure.

  • Designed and shipped a hybrid recommender system for a real estate platform — combining embedding-based and collaborative filtering approaches, served via FastAPI on AWS.

    FastAPI AWS Embeddings Collaborative Filtering Docker
  • Built a risk automation platform for a prop-tech financing workflow that flags high-risk deals, enforces underwriting rules, and generates audit logs automatically.

    Streamlit FastAPI Celery PostgreSQL Docker
  • Set up MLOps pipelines with drift detection and blue-green deployment to keep production models reliable over time.

    MLflow CI/CD Docker Redis Scrapy

Built and deployed an interactive ML education platform on Kubernetes, making complex algorithms accessible to Virginia Tech students and faculty at scale.

  • Built and deployed an interactive ML visualization platform for teaching and research at Virginia Tech — enabling hands-on exploration of ML algorithms.

    Python Dash Plotly GCP Kubernetes
  • Made complex ML concepts more accessible to students and faculty through visual, interactive exploration tools served on a production Kubernetes cluster.

    Python Flask Docker GCP

Developed a deep learning system for SCADA cyber attack detection, ranking top-3 on the BATADAL benchmark and publishing findings in the Elsevier Journal of Water Process Engineering.

  • Developed a deep learning cyber attack detection system for water distribution infrastructure — modeling spatial-temporal patterns across large-scale SCADA sensor networks in Washington D.C.

    PyTorch Graph Neural Networks SCADA Time Series Python
  • Built a production monitoring dashboard to integrate the ML pipeline into an industrial SCADA workflow, making model outputs actionable for operators.

    Python Dash Plotly SCADA
  • Validated the system against simulated poisoning attacks using synthetic data generation, achieving a top-3 ranking on the BATADAL benchmark. Published in Elsevier Journal of Water Process Engineering, Vol. 52 (2023).

    Adversarial ML Synthetic Data Research Publication

Led a 7-person team to build a satellite road quality classifier — 2nd place among 8 universities, with an 80% reduction in adversarial attack vulnerability.

  • Led a team of seven to build a satellite image classification pipeline for assessing road quality in developing regions — from raw data collection through model training and evaluation.

    TensorFlow CNN Transfer Learning Computer Vision Python
  • Improved model robustness against data poisoning attacks by developing an Autoencoder-based defense mechanism, mitigating 80% of adversarial attacks.

    Autoencoder PyTorch Adversarial ML Data Augmentation
  • Placed 2nd among eight universities in the Geo Lab Research Fellowship competition for model performance and dataset size.

    Research Team Leadership
Spare-time Builds

End-to-end agentic mobility pipeline on the Microsoft Geolife GPS dataset: classifies commute mode from raw GPS, estimates per-user carbon footprint, and simulates how Apple ESG nudges shift behavior using LLM-powered agents.

Python Streamlit LightGBM Claude API Docker Fly.io
May 2026

Built nights and weekends. Day-job production work is in Experience.

Education

Graduate research focused on machine learning systems, trustworthy AI, and scalable infrastructure. Funded by the Commonwealth Cyber Initiative (CCI).

GPA: 3.9 / 4.0  ·  Deep Learning, Reinforcement Learning, Trustworthy ML, NLP, Computer Vision, Scalable System Design.

Undergraduate foundation in computer engineering with a focus on systems, algorithms, and applied ML.

GPA: 3.7 / 4.0  ·  Machine Learning, Data Analytics & Visualization, Software Engineering, Data Structures & Algorithms, Computer Architecture, Embedded Systems.

Stack

Agentic AI

LLMs LangChain LangGraph RAG Embeddings Vector Databases Multi-Agent Orchestration

AI Tooling

Claude Code GitHub Copilot Codex Cursor Graphify Obsidian Prompt Engineering

AI / ML

TensorFlow PyTorch scikit-learn MLflow SHAP Transfer Learning Time Series Forecasting

Backend

Python SQL Go Java FastAPI Flask Celery Dash Streamlit Pandas RESTful APIs Git

Data

Snowflake Airflow MySQL PostgreSQL MSSQL Redis InfluxDB Tableau Power BI

Cloud & DevOps

Azure AWS GCP RockAI Docker Kubernetes CI/CD Jenkins Azure DevOps Grafana