Huy Nguyen

Huy Nguyen

Staff Software Engineer & AI Infrastructure

Email GitHub LinkedIn

Hello, world. I'm Huy. I write code, debug distributed systems, and occasionally break production. I have over 8 years of rigorous experience architecting scalable backend systems, training large language models (LLMs), and developing high-performance networking architectures.

I enjoy the pain of C/C++ memory management and the strict compiler checks of Rust. When I'm not optimizing server response times or tweaking CUDA kernels, I'm probably reverse-engineering an old game protocol just for fun.

root@huy-nguyen:~# cat /var/log/career_history.log

Jan 2023 - Present

Senior AI Infrastructure Engineer @ OpenAI

Working at the bleeding edge of AI hardware utilization. My primary responsibility is ensuring our GPU clusters don't melt while training next-gen Large Language Models.

  • Architected and deployed custom NCCL backends to optimize cross-node GPU communication, increasing distributed training throughput by 22%.
  • Wrote highly optimized Triton and CUDA kernels to bypass VRAM fragmentation, effectively shaving off 45% of inference latency for production APIs.
  • Collaborated with the scaling team to maintain the resilience of API endpoints handling over 50M+ requests/day.
PyTorch CUDA/Triton Python Distributed Systems
Mar 2021 - Dec 2022

Lead Backend Engineer @ TechNova Financial

Brought in to rescue a failing monolithic architecture. Led a team of 12 engineers to completely rewrite the core financial transaction engine.

  • Designed a robust, event-driven microservices architecture utilizing Rust and Go, connected via Apache Kafka.
  • Successfully handled $2M+ in daily transaction volume with a p99 latency of under 40ms.
  • Implemented zero-downtime deployment strategies via Kubernetes (EKS) and strict CI/CD GitHub Actions pipelines.
Rust Go Kafka Kubernetes
Jun 2018 - Feb 2021

Systems Edge Engineer @ Cloudflare

Worked on the core CDN and WAF routing layer. This is where I truly learned how the internet works under the hood and how to survive massive DDoS attacks.

  • Developed and optimized eBPF programs in C to drop malicious packets at the kernel level, mitigating L3/L4 DDoS attacks.
  • Maintained and scaled NGINX/OpenResty edge nodes utilizing Lua and Rust modules.
  • Automated BGP route announcements for Anycast IP failovers across 50+ global data centers.
C/C++ eBPF Lua Networking (TCP/BGP)
2015 - May 2018

Full-Stack Freelancer / Indie Hacker

The humble beginnings. Built dozens of web applications from scratch for various clients. Learned the hard way that "it works on my machine" doesn't count in production.

  • Survived the dark ages of jQuery and PHP before transitioning entirely to modern JavaScript frameworks (React/Node.js).
  • Managed my own Linux VPS servers, dealing with midnight database crashes and manual SSL certificate renewals.
JavaScript PHP MySQL Linux Admin

root@huy-nguyen:~# ls -la ./personal_projects

MMORPG Infrastructure

Knight Age Server Emulator

Reverse-engineered client protocols and built a custom server emulator in C++. Implemented raw TCP/UDP sockets via Boost.Asio to handle 5000+ concurrent players.

Cloud Utility Suite

tool.alk.xx.kg

Cloud-based web tools for developers and network administrators. Features real-time DNS propagation checkers and payload generators. Deployed on Docker Swarm.

root@huy-nguyen:~# cat ~/scripts/deploy_node.sh

#!/bin/bash
# zero-downtime rolling update for Swarm

set -e

echo "[*] Pulling latest images..."
docker pull myregistry.com/api-server:latest

echo "[*] Triggering rolling update..."
docker service update \
  --image myregistry.com/api-server:latest \
  --update-parallelism 2 \
  --update-delay 10s \
  api_cluster

if [ $? -eq 0 ]; then
    echo "[+] Deployment successful!"
    curl -X POST -H "Content-type: application/json" \
         --data '{"text":"Deployment OK"}' $SLACK_WEBHOOK
else
    echo "[-] Deployment failed! Rolling back..."
    docker service rollback api_cluster
fi

root@huy-nguyen:~# lshw -short && neofetch

Main Workstation

Custom Rig: AMD Ryzen 9 7950X, 128GB DDR5 RAM, 2x NVIDIA RTX 4090 (for local LLM training & CUDA testing).

Displays

Dual LG 27" 4K UHD monitors configured vertically and horizontally. Because reading 200 lines of logs requires vertical space.

Peripherals

Ergodox EZ Split Mechanical Keyboard with Cherry MX Brown switches. Logitech MX Master 3S mouse.

Operating System

Arch Linux (I use Arch, btw). Window Manager: i3wm. Terminal: Alacritty + Tmux. Editor: Neovim entirely configured in Lua.