cv
This is a description of the page. You can modify it in '_pages/cv.md'. You can also change or remove the top pdf download button.
Basics
| Name | Wei Zhang |
| Label | M.S. Student in Computer Science |
| zhangw2@illinois.edu | |
| Phone | (+1) 217-418-0901 |
| Url | https://en-2863.github.io/ |
| Summary | M.S. student in Computer Science at the University of Illinois Urbana-Champaign, advised by Prof. Fan Lai. Previously received B.S. in Computer Science from Peking University. Interested in machine learning systems, large language model serving, and scheduling for SLO-aware inference. |
Education
Work
-
2026.05 - Present Research Intern
Cisco
Research intern working on systems for machine learning and large language models.
-
2024.09 - 2025.04 Core Developer
Tempo: SLO-aware LLM Serving System
Built an SLO-aware LLM serving system on top of vLLM to improve responsiveness across diverse workloads.
- Implemented Just-in-Time scheduling with response-length estimation and dependency tracking, reducing SLO violations by up to 6.3×.
- Designed the Grouped Margin Goodput Maximization (GMAX) algorithm to balance per-request deadlines and batch efficiency, saving 28.5%–83.2% resources.
-
2023.10 - 2024.03 Full-stack Developer
Serverless Pilot
Developed a serverless workflow engine with DAG scheduling and pre-warming strategies under Prof. Xin Jin.
- Built backend services and frontend monitoring dashboard.
- Benchmarked performance across real workflows, improving stability and user experience.
-
2023.07 - 2023.10 System Developer
PKUJarvis
Developed a data pipeline and model-serving system for Chinese-language Q&A.
- Aggregated WeChat public account data with web crawling.
- Fine-tuned Llama2 and ChatGLM models to improve downstream accuracy.
Projects
- 2023.11 - 2024.01
Improved Open-vocabulary Segmentation with Diffusion Models
Developed a fusion module combining visual and textual embeddings, improving segmentation accuracy on open-vocabulary datasets.
- Implemented model training and inference in PyTorch.
- Evaluated prompt-engineering techniques.
- 2023.09 - 2024.01
Paint Like Painters: Image-Generation with Diffusion-based Autoregressive Framework
Implemented an autoregressive image generation framework with diffusion-based masking.
- Built training pipeline and inference engine, compared performance with one-shot diffusion baselines.
Skills
| Programming | |
| C | |
| C++ | |
| Python | |
| Java | |
| Go |
| Systems & Tools | |
| Linux | |
| Docker | |
| Kubernetes | |
| Git |
| Machine Learning & Infrastructure | |
| PyTorch | |
| vLLM | |
| CUDA | |
| Triton |
Languages
| English | |
| Fluent |
| Chinese | |
| Native |
Certificates
| TOEFL | ||
| ETS | 2025-01-01 |