EDUCATION
University of Electronic Science and Technology of China (UESTC) (Sep 2022 — Jun 2026)
University of Glasgow, Dual Degree Program (Sep 2022 — Jun 2026)
- Major: Electrical & Computer Engineering BEng; GPA: 3.87/4.0, Ranking: 2/164 (Top 1.2%).
INTERNSHIP
LLM Agent Evaluation and Post-Training Intern (Oct 2026 — Present)
Cloudnine Information Technologies Co., Ltd., LLM Engineering Department
- Background. General-purpose LLMs struggled to extract parameters from semiconductor datasheets due to limited domain knowledge.
- Approach. Adopted SFT+OPD pipeline. First converted extraction failures into targeted QA data for SFT and built a pipeline to generate agent trajectories for RSI. Then fine-tuned a Qwen2.8-27B base model after CPT on 1k+ agentic SFT examples using four NVIDIA RTX PRO 6000 GPUs, then applied OPSD using the improved model’s own feedback.
- Results. On 5,279 semiconductor datasheet parameter extraction test examples, post-training improved accuracy from 22% to 84% and recall from 36% to 82%.
RESEARCH & PROJECTS
LLMlab: A Formal-Language Testbed for Rapid RLVR Experimentation (Jun 2026)
- Motivation. The high cost of evaluating LLM algorithms slows iteration and optimization in real-world applications, motivating the development of a platform for rapid, controlled performance evaluation on synthetic data.
- Approach. Designed a deterministic formal language and built the pretraining and post-training corpora, verifier, and reward functions from scratch. Implemented a complete pipeline covering pretraining, SFT, GRPO, KD, OPD, and ran parallel ablation experiments using a 2.67M teacher and a 0.15M student. Integrated PCA projections of the loss landscape and weight trajectories, layer-wise attention heatmaps, and exposure bias metrics to analyze model behavior and training dynamics.
- Results. Enabled rapid evaluation of RLVR algorithms across task difficulty levels, achieving 100% accuracy at levels 0–3 and over 70% at levels 4–6.
LLM Post-Training: Reproducing Self-Distillation Policy Optimization (SDPO) Paper (May 2026)
- Successfully reproduced the SDPO algorithm using the verl RL framework on RunPod H100 GPUs, building on a thorough understanding of the algorithm.
- Trained Qwen2.5-3B on code generation tasks using LeetCode-style feedback (runtime errors, failed test cases) as the learning signal, and tracked 40 steps of SDPO training dynamics via WandB, logging mean reward and token-level KL divergence relative to the reference policy.
System-level Co-Design of RISCV Accelerators for Edge TinyML (Sep 2025 — Apr 2026)
Research Assistant to Prof. Yun Li, UESTC
- Motivation. Deep learning performance on edge devices is limited by power and compute constraints of embedded processors. We could design a dedicated FPGA accelerator to exploit hardware parallelism for low-power edge vision.
- Approach. Developed an NPU on an Artix-7 FPGA for real-time YOLOv8n inference. Designed a custom RISC-V instruction set extension (Xnpu) and an E2E machine learning compiler for direct hardware execution (without a soft-core processor). Implemented INT16 quantization, memory-aware instruction scheduling, parameterized RTL operators built around a 3×3 systolic array, and a fully hardware-accelerated post-processing engine (DFL and NMS). Integrated asynchronous camera and UDP video streaming with shared DDR3L memory access over AXI4, then verified the complete system with Cocotb and Bazel and validated it on the FPGA board.
- Results. Achieved real-time YOLOv8n inference on hardware and held quantized mAP within 0.3% of the PyTorch FP32 baseline. Reached peak throughput of 288 MACs per cycle and 23.4 GOPS.
YOPO: You Only Pick Once — Light Object Tracking Algorithm (Sept 2025)

- Developed a lightweight object tracking algorithm that requires only one initial selection, successfully mitigate the intense computation of DNN forward propagation on every frame.
- Utilized NCC-based matching, adaptive kernel updating, capable of tracking objects with gradual color and size changes.
Design and Visualization of a Complete Single-cycle RV32I CPU Core (Jan 2025 — Mar 2025)

Digital- Designed a single-core, single-cycle RISCV 32-bit CPU from scratch in
Verilogfor RTL simulation and inDigitalSoftware for working principle visualization, open-sourced on Github. - Built a complete datapath including PC, fetcher, decoder, register file, ALU, LRU-based L1 cache, etc., compatible with basic peripherals: GPIOs, IIC, UART, etc.
- Implemented a boot program in RISCV assembly, basic delay and GPIO libraries in
C. Compiled and simulated using RISCV GNU toolchain.
CNN/LSTM for Embedded Systems (Feb 2024 — May 2024)

- Designed and Integrated CNN and LSTM models into STM32 MCU for end-to-end patient fall detection of accuracy 95%, temperature monitoring and real-time data visualization.
- Manually collected and labeled time-series 3D acceleration dataset. Trained models on Linux, then hardcoded and accelerated them in
C++on MbedOS for real-time inference.
Human Voice Recognition Smart Car (Sept 2023 — Dec 2023)

- Designed and implemented a voice-controlled car on STM32F103 using
Cstandard libraries, supporting actions such as moving forwards/backwards, turning/sliding left/right. - Led a 4-member team in the project.
Digital Door Lock for Dormitory (Sept 2023 — Oct 2023)

- Designed and implemented an embedded digital door lock system in
C++on Nucleo L432KC MCU. - Developed basic functions include manually setting up password, automatically lock for repeated wrong passwords, OLED message displaying, etc.
- Led a 3-member team in the project.
RELEVANT SKILLS
IT Skills: Latex, Quarto Markdown, Linux, Manim, Github.
Programming: Python, PyTorch, C/C++ , Makefile.
Language: Native Chinese, Fluent English (IELTS 7).
AWARDS
Top Academic Scholarship of UESTC (Top 5%) (Dec 2023, Dec 2024)
China National Scholarship (Top 0.2%) (Dec 2024)
First Prize: 7th National College Art Exhibition and Performance (Violin section) (Sept 2024)
ACADEMIC RECORD1
| Year | Subject | Score (Full mark: 100) |
|---|---|---|
| Year 1 | Calculus I/II Linear Algebra C Programming Physics I |
91/92 84 95 88 |
| Year 2 | Physics II Signal and Systems Probability and Statistics Microelectronic Systems Embedded Processors Circuit Analysis and Design Computer Network Academic English |
96 91 92 92 95 95 94 89 |
| Year 3 | Information Theory Principles of Communication Digital Circuit Design Machine Learning Stochastic Signal Analysis Communication Circuit Design Electromagnetic Field and Microwave Technology |
91 95 86 86 82 92 88 |
See my detailed scores here.
1 I’m relatively confident in my understanding of the Boldface subjects.
INTERESTS
- Classical Music Enthusiast🎻: Concertmaster of 2nd Violin in UESTC symphony orchestra, votary of legendary composer Gustav Mahler and Johann Sebastian Bach.
- Badminton Lover🏸: Sports always refreshes me at any time.
- Learning Everything🔍: I believe everything is learnable by First Principle Thinking and curiosity.
- Volunteer Work🤝: Enjoy helping others. Over 15 hours of volunteering.