Jingyu Wang — ML Engineer & Data Scientist

Experience

Reduced request latency by ~45% via multithreading & async; raised matching accuracy from 63%→92% with multi‑strategy approach.
Designed TB‑scale e‑commerce feed ingestion with AI data enhancement; built price alerts, fake‑shop detection, context‑aware chatbot.
Led prompt engineering, LLM fine‑tuning, and safety evaluations.

Backend features & infra across GCP/AWS; Dockerized services; CI/CD with GitHub Actions.

Built GenAI text‑to‑texture tool; trained locally, deployed to Azure, integrated into production.
High‑fidelity material synthesis; PyTorch + C++/CUDA modules; integrated with ComfyUI.

Education

Sep 2021 – Sep 2024 · GPA 5.5/6 · Master Thesis: 5.75/6

Sep 2019 – Jun 2021 · Avg 18.1/20 · Rank 1/100 · Distinguished Graduate

Sep 2017 – Jun 2019 · GPA 3.97/4.00 · Rank 1/197 · National Scholarship

Skills

PythonC++CUDAShellPyTorchTensorFlowscikit-learnLangChain

PostgreSQLSparkMongoDBFirestoreGCPAWSAzure AIDockerCI/CD

LinuxGitHub ActionsOpenCVhuggingfaceSDScrumPresentationDocumentation

English (C1)French (C1)Chinese (native)German (A2)

Azure AI Engineer AssociateIBM Data Science Professional

Publications

3DV 2025 (accepted)

Relightable Information‑Sharing SDF for glossy object inverse rendering; achieves SOTA in reconstruction, material estimation & relighting.

3D VisionNeRF/SDF

SIGGRAPH Asia 2025 (accepted)

Production‑oriented pipeline for generating consistent material maps from prompts; integrated and evaluated with artists and art directors.

GraphicsDiffusion

Experience