Education Experience Research Trading Infra Skills Contact
Systems & ML Infrastructure

Eric Zhou

I build infrastructure for systems that can't go down — automated failover at Capital One, shared-autonomy research with Stanford ILIAD, and a trading system that runs unattended in production.

Scroll

Education

University of California, Los Angeles
B.A. Computer Science & Linguistics · Minor in Data Science Engineering

GPA 3.94/4.00

Sep 2024 – Jun 2028

Experience

Ironsite AI
ML Platform Engineer — San Francisco

Joining the ML platform team to build model training and serving infrastructure.

Incoming Sep 2026
Capital One — CORE Team
Software Engineer Intern, Infrastructure
  • Migrated observability into code, moving New Relic and Splunk workflows toward OpenTelemetry and Observe via Terraform.
  • Raised distributed-system resiliency with cross-region replication and automated failover paths.
  • Hardened AWS infrastructure using Kafka MSK Replicators and Aurora RDS multi-AZ patterns.
Jun 2026 – Present
Stealth Startup
Co-Founder & CTO — backed by a16z scouts
  • Led engineering and systems architecture for a proprietary in-house human attention model.
  • Owned the training and inference pipeline from data flow through production serving.
  • Kept inference reliable with Playwright coverage and real-time PostHog telemetry.
Apr 2026 – Present

Research

Systems Research Jan 2026 – Jul 2026

Programmable Software Systems Lab, UCLA — Serverless Cache for JIT Compilers

Built toward a universal serverless cache for JIT compilers to reduce cold-start latency. Extended the architecture to single-stage compilers like JAX, benchmarked against ViT/FLAX, and designed a language-agnostic profiling layer.

ICML Workshop Apr 2025 – Jun 2026

Proximal State Nudging: Reducing Skill Atrophy from AI Assistance

With Stanford ILIAD and the UCLA Robot Intelligence Lab. We introduce an algorithm that counteracts human skill atrophy under AI assistance — up to 7× larger gains in unassisted skill than standard shared autonomy, with 50% fewer collisions than unassisted practice. I built the control and simulation stack: NMPC on a kinematic bicycle model in CasADi, RL agents in Stable-Baselines3 and Gymnasium, and CARLA experiments reaching 90% autonomous parking accuracy.

Read the paper on arXiv →

Autonomous Trading Infra

PolyTrader — Prediction-Market Trading System Jan 2026 – Present

Self-built autonomous trading system that runs unattended in production: a three-bot suite provisioned with Terraform, Dockerized on a VPS, and deployed behind Cloudflare Tunnels. Captures ~5% prediction-market dislocations through mathematical modeling, Kalman filters, and regime-switching logic.

View the live dashboard →

Counter-Strike 2 (CS2) Trading Bot Jul 2024 – Dec 2025

Automated Telegram price-arbitrage bot with fuzzy matching that generated 5-figure revenue and scaled into a subscription service serving 10+ vendors. Built with Python, SQLite caching, concurrent scraping, and authenticated sessions.

Skills

Get in touch

Looking for Summer 2027 internships in systems engineering, ML infrastructure, or quantitative engineering. If something seems like a fit, reach out.