CV
Zining He's academic CV, including research and industry experience in AI systems, LLMs, and multimodal large models.
Contact Information
| Name | Zining He |
| Professional Title | Computer Science Undergraduate |
| hezining@sjtu.edu.cn | |
| Phone | +86-15662593643 |
| Website | https://xiaoheto.github.io |
Professional Summary
Computer Science undergraduate at Shanghai Jiao Tong University with experience in multimodal large-model algorithms, LLM inference, AI systems, and formal verification.
Experience
-
2026 - Present Shanghai Jiao Tong University
Undergraduate Research Assistant
Cross-Architecture Operator Differential Testing
Developed a CUDA/CANN differential testing pipeline to compare neural-network operator semantics and support GPU/NPU reliability debugging.
- Built differential testing framework for cross-architecture operator validation
- Supports GPU and NPU reliability debugging
-
2026 - 2026 SJTU Undergraduate Innovation Practice Program
Project Member
Adaptive Acceleration for MoE Large Language Models
Designed adaptive expert allocation using gate scores, layer importance, attention, and token activations for MoE inference.
- Profiled Qwen1.5-MoE and DeepSeek-V2 with PyTorch hooks and MMLU scripts
- Reduced latency by 12% on Qwen1.5-MoE prefill and 5% on DeepSeek-V2 token generation
-
2026 - 2026 Shanghai Jiao Tong University
Undergraduate Research Assistant
SpecBridge - Natural Language to Lean Specifications
NeurIPS 2026 accepted, third author. Contributed to the development and evaluation of a reconstruction-guided pipeline for translating natural-language requirements and fixed signatures into Lean specifications.
- Supported Python-based checks, Lean proof-obligation validation, and CLEVER experiments
- Project reported 6.1/15.2/28.4 percentage-point gains over few-shot + CoT baselines
-
2025 - 2025 Shanghai Jiao Tong University
Undergraduate Research Assistant
Intent-Based Test Case Generation
Converted structured and semantic testing intents into executable tests for targeted coverage and bug finding.
- Developed intent-to-test generation framework
- Improved test coverage through semantic analysis
-
Shanghai, China
Algorithm Intern, Multimodal Large Models
Shanghai Chuangzhi Academy
Algorithm internship in the multimodal large-model direction, approximately two months.
Education
-
2024 - 2028 Shanghai, China
B.S.
Shanghai Jiao Tong University
Computer Science and Technology
- Focus on AI systems, formal verification, and machine learning
- Undergraduate research in multiple projects
Skills
Programming Languages (Advanced): C++, Rust, Python, Java
AI Systems (Advanced): PyTorch, CUDA, Huawei CANN, GPU/NPU operator testing, MoE inference profiling
Formal Methods (Intermediate): Lean, specification synthesis, proof-based verification, test-based verification
Systems (Advanced): Operating systems, computer architecture, Linux, Git
Languages
Chinese : Native speaker
English : Fluent
Interests
AI Systems: Large Language Models, Model Optimization, Inference Acceleration
Formal Verification: Theorem Proving, Specification Synthesis, Program Verification
Systems Research: Operating Systems, Computer Architecture, Performance Optimization