ByteDance / TikTok
JAN 2026 — PRESENT
Staff Software Engineer · AI Infrastructure
Infrastructure owner for the Large Vision Model platform. AIGC serving, orchestration, scheduling, and model deployment with vLLM and CUDA.
Staff Software Engineer
AI Infrastructure & Distributed Systems
I own AI serving infrastructure for ByteDance / TikTok’s Large Vision Model platform, spanning model deployment, workflow orchestration, and scheduling.
Serving infrastructure for generative vision models, from research handoff to production.
Serving platform · 1,000+ QPS
vLLM · CUDA · ComfyUI
I own workflow orchestration and scheduling, with a focus on improving GPU utilization.
For an asynchronous AIGC serving framework, I make architecture decisions, coordinate work across teams, and work with users and leadership on adoption and project funding.
I deploy and serve models including Seedance and Seedream, alongside open-weight models such as Qwen and MiniMax H3.
I mentor engineers and help teams adopt coding agent workflows. I also use agents to automate repetitive engineering work.
JAN 2026 — PRESENT
Infrastructure owner for the Large Vision Model platform. AIGC serving, orchestration, scheduling, and model deployment with vLLM and CUDA.
APR 2024 — DEC 2025
Led a cross-functional CVV initiative for Reality Labs purchases. Built payment risk systems for Reality Labs’ app store, including fraud detection, real-time anomaly detection, and rate limiting.
APR 2017 — MAR 2024
Architected foundational SOA services for Airbnb’s move from a monolith to microservices, built a Kafka-based pub-sub system, and managed critical services including Dynamic Configuration.
Led the integration of ChatGPT into Airbnb’s customer support chatbot platform. Also worked across search, advertising, and recommendations.
FEB 2016 — APR 2017
Helped build and launch an internal insurance platform.
JUN 2015 — AUG 2015
Designed and built a web-based tool to automate log analysis for Amazon EMR.
I’ve architected foundational services at Airbnb and led a cross-functional CVV initiative at Meta. Today I own AI serving infrastructure at ByteDance / TikTok.
The common thread is production engineering: how services are deployed, coordinated, and kept reliable at scale.
Outside work, I study model serving with vLLM and SGLang, including continuous batching and KV caching. I also practice operator fusion with OpenAI’s Triton and explore CUDA Graphs.
San Mateo, California
USC · M.S. in Computer Science
vLLM · CUDA · PyTorch · Triton
For opportunities in AI infrastructure, model serving, or distributed systems, connect with me on LinkedIn.