Eric Suen

Senior platform engineer · Owned deployment infrastructure for ~1,000 engineers at Block · Go, Kubernetes, distributed systems

Currently exploring senior backend and platform engineering roles in SF and remote.

What I've been building lately

Launchpad Go, Kubernetes, Helm, GitHub Actions, Postgres, Prometheus, Grafana

A deployment tracking platform built to mirror what a strong startup runs in production. A Go API backed by Postgres, deployed to Kubernetes via Helm with GitOps through Argo CD, automated CI/CD, HTTPS ingress, and full observability through Prometheus and Grafana. Features environment separation (staging/production) with declarative deployments where git is the single source of truth. See the project →

Astral Rift Unreal Engine 5, C++, HLSL, Python, JAX

Your shields are your weapons. A multiplayer space combat game where every trigger pull costs you the defenses that keep you alive, built entirely in C++ on Unreal Engine 5. Eight ships across five factions, 20 abilities, 5 game modes, and neural network bots trained via reinforcement learning across 50 billion+ steps—with automated evaluation that diagnoses behavioral gaps and tunes reward shaping across 23 training phases. Client-side prediction with server reconciliation and lag compensation. 375+ automated tests. See the project →

Barbell 5x5 Swift, SwiftUI, iOS, watchOS

An app that knows what workout you should do today, tells you exactly what weight to load, and gets out of the way. 8 built-in programs, HealthKit readiness scoring, Apple Watch companion, and 50+ exercises with animated form guides. See the project →

Pickleball Rally Clipper Python, YOLOv8, OpenCV

I play a lot of pickleball and got tired of scrubbing through hours of match footage to find key rallies. I'm building a computer vision pipeline that cuts dead time from recordings, identifies shot types (dinks, drives, drops, lobs, overheads, counters), measures ball velocity, and uses pose estimation to compare player form against professional technique. Built on a custom-trained YOLOv8 model with YOLOv8-Pose for joint tracking. In development—demo coming soon.