Hello Friend, Welcome to My Blog
I am Ye Wentao, currently a Master of Computer Science student at Cornell University, expected to graduate in May 2025.
I am actively seeking full-time opportunities in SDE / MLE / AI Engineer. If you are interested, please reach out to me!
Email: zhyanwentao@outlook.com
Summary
Software engineer with 2+ years of experience and 1.5+ years of internships. Experienced in developing and optimizing large-scale AI frameworks with Python and C++. Demonstrated ability in open-source contributions and technical innovation as a co-founder of two tech-startups. Passionate about learning and applying new technologies in fast-paced environments.
Experience
SenseTime
Deep Learning System Development Engineer (SenseCore) | Shanghai, China
Jul 2022 - Aug 2024
- Developed LazyLLM, a large language model application framework, including the construction of RAG system for more accurate context-aware suggestions, and a parallel mechanism to enhance large-scale deployment efficiency. GitHub
- Designed and built an Op Inferrer mechanism for the DeepLink Framework, constructed a CI/CD pipeline for traditional model accuracy verification. GitHub
- Implemented multi-chip operator adaptations and optimized operator performance for DIOPI (Device-Independent Operator Interface), resulting in a 20% performance improvement. GitHub
- Stabilized Llama2 70B training across 64 Huawei 910B cards, boosting tgs (tokens per GPU per second) from 10% to 70% compared to A100.
- Developed and patented the One-iter Tool (CN117312173A) for rapid model accuracy comparison, enabling correctness validation within minutes
SenseTime - R&D Intern
Research Institute (Deep Learning Frameworks) | Beijing, China
Jan 2021 - Jul 2022
- Developed Inspector service for deep learning training data collection, meeting commercial delivery standards.
- Reconstructed PAVI, a training data collection tool, covering batch processing, concurrency, caching and retransmission, ensuring stable and reliable data uploads and improving efficiency by tens of times.
- Invented the function-level task scheduling tool Auction, obtained a patent (CN115033366A) and enabled distributed training.
- Proactively expanded business channels by integrating the task scheduling platform, data, and deep learning model teams, successfully completing the automatic model annotation project.
Wuhan Hongyuan Investment & Creation Technology Services Co., Ltd.
Co-founder | Wuhan, China
Nov 2019 - Sep 2020
- Led development of software and hardware solutions for the Henan Shangqiu City Procuratorate AI Exhibition Hall.
Yuye Tech (Wuhan) Development Co., Ltd.
Co-founder | Wuhan, China
Jun 2019 - Nov 2019
- Led the architecture and service development of Dingdong Baotuan, a competition teaming platform for college students. Successfully launched and gathered hundreds of users.
Education
Cornell Tech at Cornell University | New York, USA
Master of Computer Science
Aug 2024 - May 2025
Wuhan University | Wuhan, China
Bachelor of Software Engineering (Excellent Engineer Program)
Sep 2018 - Jun 2022
- Selected for the National Excellent Engineer Training Program (top 5% acceptance rate)
- Cumulative GPA: 3.91/4.0, core course GPA: 3.92/4.0
- Technical Director and Vice President of the Wuhan University Microsoft Student Club
Projects
PyTorch Contribution | May 2023 - Present
- Authored a series of blog posts analyzing source code and the internal mechanisms of PyTorch, including Tensor Storage, CPU & Cuda Operators, Dispatcher, TensorIterator, Autograd, etc. Blog
- Made multiple contributions to the PyTorch community, including optimizing Cudnn operator performance, which resulted in a 15% efficiency increase. GitHub
MicroTorch | Jun 2023 - Present
- Built a simplified PyTorch implementation from scratch for educational purposes. GitHub
- Including Tensor implementation, memory management, forward/backward computation of CPU and CUDA operators, computational graph, and automatic differentiation.
- Utilized pybind to export C++ to Python, enabling pip installation and unit test-driven development.
Other Open-Source Projects
- You After Taking Drugs: Face Change Simulation System: GitHub
- CMU CSAPP-15213 course notes and labs: GitHub
- TinyNN: Simplified implementation of FCNN and CNN: GitHub
- Sicpy: Typeless programming language Compiler: GitHub
Competitions
- 2021 China University Computer Contest - WeChat Big Data Challenge - National Third Prize (Top 2% / 6000+ teams)
- 2020 China University Computer Contest - Huawei Big Data Challenge - National Third Prize (Top 2% / 1400+ teams)
- 2020 China Service Outsourcing Innovation and Entrepreneurship Competition - National Second Prize (Top 2%)
- 2019 China University Computer Capability Challenge - National First Prize (Runner-up / 400+ teams)