Experience

Senior Machine Learning Engineer - LLMs

TikTok

  • Lead the development of Model-as-a-Service infrastructure and high-performance LLM serving systems, integrating vLLM and SGLang across model runtimes, distributed execution, and production deployment workflows.
  • Design model-system co-optimizations for scheduling, continuous batching, KV-cache and memory management, observability, fault isolation, and reliability to improve the efficiency and operability of large-scale inference.
  • Co-authored KVDirect on distributed disaggregated LLM inference and developed CommBridge, a portable cross-GPU communication runtime spanning KV-cache transfer, MoE communication, and heterogeneous RDMA backends.
  • Work across inference algorithms, runtime design, networking, and hardware utilization to make foundation-model experimentation faster and production serving more dependable.

Quantitative Algo Developer - Smart Contract (DeFi 2.0)

Aperture Finance

  • Served as technical owner for Solana quantitative-strategy projects and designed a pseudo-delta-neutral hedging strategy using leveraged liquidity-farming protocols.
  • Built a research pipeline that combined autoencoder-based risk-factor extraction, GAN-generated synthetic data, wavelet denoising, and Kalman filtering over five years of historical market data.
  • Evaluated individual and composite alpha factors through offline backtesting with Zipline and Backtrader, reducing estimated deployment risk for candidate strategies by 30%.
  • Implemented Rust smart contracts that converted model-generated signals into dynamic long-short portfolio adjustments; the deployed strategy generated $2M in net income in one month.

Staff Engineer - Cloud Infrastructure | Tech Lead

Binance

  • Led an eight-engineer team to design, build, and launch Themis, Binance’s first global strategy-distribution platform for traffic allocation and predicate-based decisioning.
  • Established the platform’s offline simulation, production-data feedback, and experimentation workflows, supporting up to 1,000 concurrent A/B experiments and 20,000 strategy releases per day.
  • Designed the online-offline evaluation loop used to improve resource-allocation decisions across products serving 200M users and approximately $1B in weekly transaction volume.
  • Owned technical direction and delivery across the platform, deployment tooling, administration workflows, and client integration.

Cloud Infrastructure Staff Engineer | Tech Lead

Alibaba Cloud

  • Led a five-engineer team responsible for high-scale personalization and messaging infrastructure during two Tmall Double 11 campaigns, including the 2019 event that processed RMB 268.4B in daily GMV.
  • Designed a behavior-aware advertising and promotion-ranking system using purchase history, search activity, customer profiles, and engagement signals; the program was associated with 48% annual revenue growth and an 8% increase in purchase rate.
  • Improved online and offline recommendation pipelines for push, banner, and SMS campaigns by 25%; expanded user-profile features and raised XGBoost/logistic-regression model AUC from 0.59 to 0.77.
  • Architected the Taco distributed messaging ecosystem to process more than 1.1B notifications per day with over 97% delivery, 0.72-1.2 second latency, 750K peak connections, and 30K-50K QPS.

Infrastructure Team Lead

Shanghai Function Internet Finance Information Service Co., Ltd.

  • Led the engineering team that built a Java-based infrastructure snapshot and caching system for read-heavy services, reducing average response time by 65% and improving robustness by 81% at petabyte-scale daily cache volume.
  • Directed the migration of 12 containerized backend services to Alibaba Cloud Kubernetes and redesigned the Redis/MySQL data layer, improving measured platform stability by 65%.
  • Designed an intelligent investor-project matching system combining recommendation, text summarization, and text-to-speech components, increasing successful project bids by 40%.

Infrastructure Team Lead

PING AN

  • Designed the 1QB Wallet recommendation system to match customers with financial products using behavioral and search signals, improving recommendation accuracy by 30%.
  • Built a Go and Kafka event-driven transaction system for millisecond-scale promotional ordering by millions of users; the platform contributed RMB 10B in incremental GMV during its first year.
  • Re-architected distributed task processing with Celery, RabbitMQ, and MySQL, reducing queue latency by 20%, increasing throughput by 4x, and lowering memory consumption by 40%.
  • Established code review, static analysis, package deployment, backup distribution, and CI/CD workflows, improving engineering delivery efficiency by 5x.

Research Engineer

Quatanium Technology Co., Ltd.

  • Designed and built the company’s first cluster observability platform, processing millions of logs per day through Logstash, Kibana, and Grafana with 99.999% annual availability.
  • Developed containerized CI/CD pipelines for AWS that automated build, packaging, and production deployment for an average of 200 integrations per day.
  • Built an automated error-detection and ticketing service for operations teams, reducing incident-response waiting time by 80%.

Education

Master of Information and Data Science, GPA 3.966/4.0

University of California, Berkeley, CA, United States of America

Master of Science in Computer Science, GPA 4.0/4.0

Stanford University, CA, United States of America

Master of Science in Software Engineering, GPA 3.95/4.0

Carnegie Mellon University, CA, United States of America

Bachelor of Engineering in Computer Science, GPA 3.8/4.0

Huzhou Normal University, Zhejiang, China

Skills
Technical Skills
golang Golang
Docker Docker
MySQL MySQL
PostgreSQL PostgreSQL
Nginx Nginx
RabbitMQ RabbitMQ
Redis Redis
jenkins Jenkins
JS JavaScript
H5 HTML5
css3 CSS3
vue Vue
weex Weex
electron Electron
apple Apple
AI Adobe Illustrator
PS Adobe Photoshop
sketch Sketch
git Git
gitlab Gitlab
phabricator Phabricator
Accomplishments
Preparing for Google Cloud Certification: Cloud Architect Specialization
Google Cloud ∙ June 2021
Architecting with Google Compute Engine Specialization
Google Cloud ∙ June 2021
Architecting with Google Kubernetes Engine Specialization
Google Cloud ∙ May 2021
Programming with Google Go
University of California, Irvine ∙ May 2021
ACM Professional Membership
ACM, Association for Computing Machinery ∙ July 2020
IEEE Professional Membership
IEEE ∙ July 2020
IEEE-CS Professional Membership
IEEE Computer Society ∙ July 2020
CCF Professional Membership
China Computer Federation ∙ July 2020
Apple Teacher
Apple ∙ June 2020
AI For Everyone
DeepLearning AI ∙ June 2020
DeepLearning.AI TensorFlow Developer Specialization
DeepLearning AI ∙ June 2019
Deep Learning Specialization
DeepLearning AI ∙ June 2019
Machine Learning
Stanford Online ∙ March 2018
The 10 Best Original Authors of 2016 juejin.im
juejin.im ∙ December 2016
The 2016 Best Recommended Author of jianshu
jianshu.com ∙ December 2016
Network Engineer
Ministry of Human Resources and Social Security Administration ∙ May 2012
National Computer Rank Examination (Level 4 Network Engineer)
Ministry of Education ∙ March 2012
Software Designer
Ministry of Human Resources and Social Security Administration ∙ November 2011
The 36th ACM-ICPC Asia Regional Contest Shanghai Site 2011 Fudan University - Honorable Mention
ACM-ICPC International Collegiate Programming Contest ∙ October 2011
The 8th Zhejiang Province University Programming Contest - Bronze Medal
Ministry of Education ∙ June 2011
The ACM-ICPC Asia Regional Google Cup 2011 ACM-ICPC China Shanghai Fudan Invitational Programming Contest Fudan University - Bronze Medal
ACM-ICPC International Collegiate Programming Contest ∙ May 2011