CV

Education

2024.11 ~ 2025.7: LLM Algorithm Researcher
- Shanghai AILab.
- I developed a Hungarian low-resource LLM fine-tuning pipeline based on InternLM2.5 7B and Qwen2.5 7B by conducting post-pre-training with 20B Hungarian corpus and generating filtered high-diversity supervised data, and the optimized model surpassed GPT-4o mini on all 27 cross-domain benchmarks.
- I proposed IDEAL, an influence function-driven automatic data proportion adjustment framework for multi-capability LLM alignment to balance model performance across diverse tasks, which achieved a 7% comprehensive performance gain over competing methods and became my sole-first-author work accepted by ICLR 2026.
Summer 2022: Researcher
- Shenzhen Flintstone Robotics Corporation Limited.
- Duties included: Control and Design of Robot Arm.
Fall 2021: Researcher
- The Beijing General Artificial Intelligence Research Institute.
- Duties included: Design and implement a surface robots.

DL(Pytorch)
RL
LLM
Embedded Development
- C Development
- C++ Development
- Robot Development(ROS, Python)

The 24th “Electronic Design Contest” held at Tsinghua University. 2023.4
- Received the highest honor, the Grand Prize.
The Baidu Creative Team participating in the National University Smart Car Competition: Silk Road. 2022.8
- Primarily responsible for the lane tracking and vehicle control aspects, respectively accomplishing the learning methods for traditional computer vision-based lane tracking and autonomous lane tracking using neural networks. Achieved the first prize in the national finals of the competition.
RoboMaster University Sim2Real Challenge.2022.5
- The inaugural ICRA Sim2Real Robot Challenge, a simulated-to-real-world intelligent vehicle competition jointly organized by DJI and AIR. Primarily responsible for obstacle avoidance path planning in navigation SLAM. Achieved the 1st Prize, securing the second place.
Tsinghua University’s “Xin Dong Plan” Science and Technology Winter Camp. 2019.12
- Accomplished autonomous line tracking and obstacle avoidance for the vehicle, successfully achieved image recognition (including traffic lights), ultimately receiving the highest honor, the Grand Prize, in the Xin Dong Plan.