SDE Intern at Optum · UBC CS ’27

I’m David Zhao,I build the platformsthat serve LLMs.

I’m a Computer Science student at UBC and a Software Development Engineer co-op at Optum — a year-plus of Kubernetes, CI/CD, and GCP in production healthcare software. This year I’m going one layer down the stack: LLM inference and serving.

EmailGitHubLinkedIn

Kubernetes GCP Go PyTorch vLLM OpenTelemetry

Impressionist water study in warm bone tones with rust water lilies, after Monet
Water study no. 21 · after Monet

01Profile

About

Vancouver, Canada

I’m studying Computer Science at the University of British Columbia, expected to graduate in 2027. In parallel, I’m building software in industry at Optum (UnitedHealth Group), working on the platform behind an enterprise healthcare SaaS — microservices on GKE, CI/CD, and release tooling.

I’m strongest where backend systems meet operations, and I’m spending the next year going one layer down into LLM serving: a serving gateway on GKE now, open-source contributions this fall, and an inference engine written from scratch over the winter. Everything ships with benchmarks.

Education
UBC Computer Science, B.Sc. 2027
Current
SDE Intern at Optum
Focus
LLM inference & serving, Kubernetes platforms

02Work

Experience

May 2025 – Present

Optum · Enterprise Imaging

Software Development Engineer Intern

Optum (UnitedHealth Group) · Richmond, BC

Optum logo
  1. 01Develop and maintain 15+ microservices on a GCP-based SaaS platform using C#/.NET and Node.js.
  2. 02Architect CI/CD workflows with Jenkins and GitHub Actions to make deployments more consistent across environments.
  3. 03Build end-to-end automation with Playwright and Selenium to expand coverage and speed up release cycles.
  4. 04Debug production-level issues across Docker, Kubernetes (GKE), shell tooling, and application code.

.NET/C# Node.js GCP GKE Docker Jenkins GitHub Actions Playwright

03Edge

Strengths

The through-line in my work: systems should be measurable — deployable, observable, benchmarked — whether that’s a healthcare platform or an inference server.

01

LLM serving infrastructure

My current build focus: the layer between models and users — gateways, routing, and cache-aware scheduling on Kubernetes, starting with Relay, a Go gateway in front of vLLM on GKE.

02

Platform engineering

The foundation: co-op terms at Optum running microservices, CI/CD pipelines, and GKE deployments for production healthcare software, where reliability is not optional.

03

Measurement discipline

Claims come with numbers or they don’t ship: latency profiles (p95, time-to-first-token), load replays of real traces, and A/B comparisons against honest baselines.

04Toolkit

Skills

The stack I run in production today — plus the serving-layer stack I’m building with in public this year.

Languages

TypeScript, JavaScript, C#, Python, SQL

Backend

.NET, Node.js, REST APIs, PostgreSQL, Microservices

Cloud & Platform

GCP, Kubernetes (GKE), Docker, Jenkins, GitHub Actions, CI/CD

Testing & Quality

Playwright, End-to-end automation, Performance testing

Current focus

Go, PyTorch, vLLM, OpenTelemetry, Inference benchmarking (TTFT, throughput)

05Contact

Let’s connect

I’m always open to talking about LLM infrastructure, platform engineering, internships, UBC, or interesting systems problems.

dykzhao@gmail.com