Newton (Nick) Duong

AI Architect · 20+ Years in Enterprise Systems & AI Solutions
duong.nick@gmail.com | 860-729-4235 | duong.casa | LinkedIn | GitHub | Middletown, CT | US Citizen

Summary

Enterprise AI architect with deep expertise in generative AI systems, cloud-native architecture, and the full ML model lifecycle. Recent work includes RAG/GraphRAG architectures with vector databases, multi-agent AI systems using GCP ADK with Agent-to-Agent patterns, and LLM-powered security analysis platforms. Skilled at defining integration patterns across LLMs, vector databases, APIs, microservices, and event-driven workflows. Technical leadership spans architecture governance, cross-functional design reviews, and mentoring engineering teams. Committed to responsible AI, data governance, and security in production-ready solutions.

Technical Skills

Languages
Python, Java, TypeScript, JavaScript
Cloud
Google Cloud Platform, Microsoft Azure, Amazon Web Services
GenAI / LLMs
Claude (Opus/Sonnet), Claude Code, Gemini 3.0, ChatGPT, Llama, Prompt Engineering
Agents / RAG
LangChain, LangGraph, GCP ADK (A2A), RAG/GraphRAG, ChromaDB, Neo4j, FAISS
ML / MLOps
PyTorch, Vertex AI, MLflow, LLMOps
Frameworks
Spring Boot, React, Node.js, FastAPI, REST APIs, Microservices, IBM API Connect
DevOps
GitHub Actions, Jenkins, AWS CodePipeline, Docker, Kubernetes/OpenShift, Terraform
Databases
DuckDB, Oracle, PostgreSQL, MongoDB, Redis, SQLite

Experience

Senior Staff Software Engineer Jul 2025 – Present
The Hartford Insurance Group, Hartford, CT
ADK-GitHubScanner – AI-Powered Security Analysis Platform
  • Architected end-to-end multi-agent AI system using Google ADK with Gemini 3.0 for enterprise security vulnerability detection
  • Designed Agent-to-Agent (A2A) orchestration with 3 specialized agents running in parallel via asyncio
  • Implemented pure LLM-based security analysis replacing traditional SAST tools, achieving deterministic results (σ<5 score variance)
  • Architected model-serving infrastructure optimized for latency and throughput using Vertex AI
  • Designed cloud-native deployment on GCP Cloud Run with GCS storage and auto-scaling
  • Established architectural standards including MAX_CAP deep analysis protocol with hallucination detection (<5% threshold)
  • Deployed infrastructure using Terraform for GCP Cloud Run and associated cloud resources
  • Implemented CI/CD pipelines using AWS CodePipeline for automated testing and deployment
  • Mentored junior developers on AI best practices, prompt engineering, and responsible AI
Staff Software Development Engineer Jun 2007 – Jul 2025
CVSHealth / Aetna, Inc., Hartford, CT
Retrieval-Augmented Generation (RAG/GraphRAG) Architecture
  • Led end-to-end architecture design for RAG/GraphRAG systems serving enterprise Data Science teams
  • Architected integration patterns connecting LLMs to vector databases (ChromaDB) and graph databases (Neo4j)
  • Implemented scalable RAG pipelines using Python, LangChain, LangGraph, and Vertex AI
  • Designed containerized solutions deployed to GCP Cloud Run via GitHub Actions CI/CD
  • Created A/B testing dashboard architecture to measure GraphRAG performance metrics
Enterprise Framework Modernization – AI-Assisted Architecture
  • Led cross-functional initiative to modernize legacy Spring applications to Spring Boot
  • Designed automated refactoring patterns using OpenRewrite and Claude Code for enterprise-scale migration
  • Created reusable recipes and architectural standards adopted across multiple departments
  • Facilitated architecture reviews and documented decision records for governance
Cloud-Native & Container Architecture
  • Architected Platform-as-a-Service solution using Docker and Kubernetes/OpenShift
  • Designed cloud ML platform patterns for Spring Boot applications in containerized environments
  • Established CI/CD architectural standards using Jenkins, Nexus, and UDeploy
  • Developed REST API architectures with Redis caching for high-throughput chat services
API & Microservices Architecture
  • Led architecture governance for API management using CA Layer7 and IBM API Connect with OAuth2
  • Designed integration patterns for mobile (iOS/Android) and web applications
  • Implemented enterprise SSO architecture with Siteminder
  • Created monitoring and observability dashboards using Splunk
Aetna Enterprise Framework (AEFW)
  • Developed and maintained enterprise-wide Spring Framework used across the organization
  • Created reference architectures and KickStart applications for rapid project onboarding
  • Provided technical mentorship and support to project teams on architecture and deployment
Staff Engineer / Technical Team Lead Apr 2005 – Jun 2007
Pratt & Whitney, UTC, East Hartford, CT
Advance Diagnostics Engine Management System (Phoenix)
  • Designed and implemented enterprise system for airplane fleet engine management
  • Built multiple web applications for diagnostics and fleet engine forecasting
  • Evolved a robust Services Oriented Architecture using Spring, Tapestry, and Hibernate
Flight Acquisition System Transmission (FAST)
  • Developed system using XML, Digital Signature/Encryption
  • Created BIRT reports accessing Oracle database; built rapid reporting via JSP/Hibernate
Senior Java Developer Oct 2002 – Apr 2005
Aetna, Inc., Hartford, CT
Enterprise Architecture Integration (EAI)
  • Built middleware infrastructure for EAI components using MQSeries
  • Supervised offshore and junior developers in EAI solutions and framework
Enterprise Portal
  • Designed and implemented portal system following Sun J2EE best practices

Education & Certifications

M.S. Computer Information Systems

Boston University
2007

B.S. Information Systems

University of Connecticut
1998

Sun Certified Java Programmer

Java 2 Platform
2000

Personal Projects

MISO Energy Trading Platform

  • Architected end-to-end multi-agent AI system for MISO energy market analysis with 541+ Python files across 6 core projects
  • Designed DAG-based workflow orchestration with LLM routing supporting multiple providers (Anthropic Claude, Ollama)
  • Implemented PyTorch LSTM-based ML models for congestion forecasting with MLflow experiment tracking
  • Built hybrid ML training infrastructure: Mac Mini M4 with MPS GPU acceleration and Dell Xeon server for large batch training
  • Built graph analysis system using DuckDB with Property Graph Query extension (2,546 vertices, 623K edges)
  • Architected MCP server for Claude Code integration enabling natural language database queries
  • Developed production FastAPI web dashboard with Plotly.js visualizations and multi-ISO support
  • Achieved 70x optimization on async data collection pipelines (38 seconds per day)