Huasong Shan

About Huasong Shan

15+ years’ experience in the internet (e-commercial), autonomous vehicles, and chip companies, good command of various technical domains: large-scale distributed systems, cloud computing, AIOps, MLOps for various AI applications.2. Open source contributions: ContainerDNS on Kubenetes:https://github.com/tiglabs/containerdnsFederated Learning: https://github.com/cyqclark/fedlearn-algo3. 10+ technical patents worldwide and 10+ top-tier publications in computer system/security/AI domains; regularly serve as TPC for ACM & computer conference (e.g, EuroSys, SoCC, USENIX Security, ICDCS, & hl=en)My focus engineering/research area:1. Build highly available, scalable, and performance system(micro-services, parallelized pipeline)2. Cloud computing: resource management, performance analysis, abnormal detection, root-cause diagnosis, etc.3. MLOps for various AI applications: e-commercial advertisement text generation, news recommendation, perception model validation for autonomous vehicles4. CI/CD: build and test automationTechnical Skills: AWS/Azure, Docker/Kubernetes, MongoDB, RabbitMQ/Kafka, Redis, Jenkins, Git, Redash, Python, Java, C/C++, Make, SQL, NumPy, Pandas, Django, Flink/Spark, Airflow, MLflow, Pytorch.

Experience

1. Senior Staff Engineer

   XPENG

   Jul 2022 — Present · Santa Clara County, CA, US

   CI/CD and MLOps, ML Platform: On-target Test for Autonomous Vehicle (MongoDB, RabbitMQ/Kafka, Jenkins, Redash)(1) Large-scale dataset on-target model evaluation/serving/inference for Perception model (100s of thousands of jobs per day)(2) Dataset ingestion pipeline: transcoding, validation, DDS replay/record (TB per day)(3) Model formate transition to TensorRT (10s of thousands of jobs per day)(4) Autonomous Vehicle Full Stack Performance Evaluation Test2. Cloud Resource Management: reliability and availability On-premise ARM/GPU cluster management and reliability, node anomaly detection, and auto-healing on K8S for NVIDIA Orin cluster (Kubernetes, Docker)3. Workload Orchestrator: performance and scalabilityEfficiency optimization, reducing the overhead of the on-target test, via batch scheduling algorithm and parallelized pipeline

Education

• Huazhong University of Science and Technology

  Master's degree, Computer Science

  2003 — 2006

• Huazhong University of Science and Technology

  Bachelor’s Degree, Computer Science

  1999 — 2003

• Louisiana State University

  Doctor of Philosophy - PhD, Computer Science

  2015