Skip to content
Simon Xu

Software Engineer, AI/ML

Simon Xu

I build AI products end to end, from the model to the interface.

Portfolio AILos Angeles, CA, USAOpen to new opportunities
See projectsResume

Enter sends the suggestion as-is · ⌘K anywhere

01 — About

About

I fine-tune and evaluate models, then ship them inside systems people can trust: evals that block a bad release, a gateway other teams can call without thinking, retrieval that cites its sources, runtimes that pick up where they crashed, and a UI that gets used every day. Right now I'm looking for Applied AI, Full-Stack AI, or Forward Deployed Engineer roles where that whole loop is my job.

Software engineer on Highmark's AI/ML team, building AI tools employees actually use: a call-center training platform where an LLM plays the customer on a live call and a rubric-based judge, tuned against coach-scored transcripts, grades the roleplay; the internal LLM gateway; and a multimodal assistant serving 30K users and 6M+ prompts a year.

Outside work: Proxy Loop (fine-tuned Qwen3-8B, held-out task completion 58%→67%) and Conductor (a Go runtime for agent workflows that crash and resume without redoing side effects).

Simon Xu

Software Engineer, AI/ML

Status
Open to new opportunities · relocation OK
Location
Los Angeles, CA, USA
Education
Master of Science in Computational Science & Engineering Georgia Institute of Technology · December 2024
Focus
Applied AI Engineer (LLM products, agents, evals)Full-Stack AI Engineer (model to UI)Forward Deployed Engineer (shipping AI with customers on-site or embedded)Post-training and evals (SFT, QLoRA, LLM-as-judge)
studio pass · @SimondXu

Currently into

  • Post-training small models (Qwen3-8B, QLoRA)
  • Temporal workflows that wait days for a human
  • Go, for runtime work
  • LLM-as-judge calibration
  • Agentic RL
  • Claude Code and Codex as daily drivers
  • A 112-track playlist I started in 2022

02 — Experience

Experience

Aug 2024 - Present
Remote, USA

Software Engineer (AI/ML team) · Highmark Inc.

Call-center voice training with an LLM on the other end of the line, the judge that grades it, and the registry and evals that decide which agents reach production.

  • Shipped a real-time AI call-center training platform (React/FastAPI/WebSockets) where an LLM plays trainer-configured customer personas over the phone, replacing $1M+/year of consultant-led training.
  • Designed the rubric-based LLM-as-judge that scores each completed roleplay, tuning it against coach-scored transcripts until its scores tracked the coaches' before trainees saw any feedback.
  • Built an enterprise AI-agent lifecycle control plane governing versioned releases from staging to production via a centralized registry, durable evaluation runs, policy-gated GitLab promotions, and cross-environment observability.

May 2024 - Aug 2024
Remote, USA

Software Engineer Intern (AI/ML team) · Highmark Inc.

The internal LLM gateway, a policy Q&A bot that cites its sources, and the multimodal assistant that became the company's GenAI platform.

  • Developed a Retrieval-Augmented Generation system using Flask, integrating embedding models and Gemini to answer employee queries on internal policies with cited sources; surveyed users reported 70% less search time.
  • Engineered the internal LLM API gateway exposing a single Gemini endpoint to engineering teams company-wide, with schema-validated JSON output, per-consumer cost attribution, and circuit-breaking on provider errors.
  • Built and scaled an internal multimodal AI assistant on Vertex AI (Gemini), enabling employees to process text, PDFs, images, and voice through a React and FastAPI stack; serves 30K users and processes 6M+ prompts annually.

03 — Selected work

Selected work

Concept artwork

Proxy Loop

A fine-tuned Qwen3-8B agent with a Fast/Slow harness and Temporal-backed durable workflows that act for the user across phone, browser, and email.

  • Python
  • Qwen3-8B
  • QLoRA
  • PEFT
Concept artwork

Conductor

A Kubernetes-native Go runtime for crash-resumable LLM workflows with event-sourced recovery, PostgreSQL leases and fencing tokens, and OpenTelemetry observability.

  • Go
  • Kubernetes
  • PostgreSQL
  • Event Sourcing
Concept artwork

Engram

A Python MCP memory harness for carrying decisions, project facts, and preferences across AI coding sessions while detecting stale context.

  • Python
  • MCP
  • Event-Driven Architecture
  • Agent Memory
Concept artwork

FigBrain

An MCP-powered Figma agent that uses a LangGraph supervisor and ReAct tool loops to turn natural-language prompts into FigJam components.

  • Python
  • TypeScript
  • LangGraph
  • MCP

↵ ask · esc close

Keyboard shortcuts

⌘ K / Ctrl K
Ask the portfolio AI
/
Ask the portfolio AI
g h
Go to Home
g p
Go to Projects
g r
Go to Resume
g a
Go to Ask
g m
Go to Music
t
Toggle light / dark
?
This sheet
h i
Say hi