About Me
我是钱一略，目前是加州大学圣地亚哥分校的一名硕士生，主修机器学习和数据科学.
我于2024年从北京大学毕业，获得人工智能学士学位。在北大期间，我是元培学院首届通班（通用人工智能试验班）的成员，并与人工智能研究院有深入合作。
我的学术和职业兴趣包括但不限于：
- 学术方向： 机器学习、（机器与人类的）认知推理、大语言模型、机器人规划。
- 工业方向： 自动驾驶、推荐系统、云计算/分布式计算、数据驱动学习、软件开发（特别是与人工智能和人工智能产品相关的领域）
欢迎随时通过 [Github] [领英] [电子邮件] 与我联系，我很乐意分享想法和信息。
Education
University of California, San Diego (加州大学圣地亚哥分校)
Master of Science in Machine Learning and Data Science | Sep 2024 - Spring 2026 (Expected)
- Department of Electrical and Computer Engineering
Peking University (北京大学)
Bachelor of Engineering in Artificial Intelligence (Computer Science) | Sep 2020 - Jul 2024
- Yuanpei College
- Member of the first Tong Class
Professional Experience
WeRide (文远知行)
Software Development Engineer Intern (Perception Group) | San Jose, CA | Jun 2025 - Sep 2025
- Working on the onboard detection model
- Enhancing the trajectory prediction model to generate diverse, multi-modal forecasts, reducing indidents by 5% in simulation
Megvii Technology Limited (旷视科技)
Software Development Intern (Self-Driving) | Beijing, CN | Jun 2023 - Dec 2023
- Models and tools development for self-driving vehicles
- Proposed Transformer neural networks to simulate vehicle trajectories
- Developed visualization tools to better understand vehicle behaviors
- Built benchmarks to evaluate trajectories
Beijing Institute for General Artificial Intelligence (BIGAI) (北京通用人工智能研究院)
Research Intern | Beijing, CN | Mar 2022 - Jun 2023
- Individual Research in Cognitive Reasoning Lab, under supervision of Prof. Wei Wang and Dr. Lifeng Fan
- Project on concept disentanglement, causal transition, symbolic reasoning, and visual planning
- Project on 3-D spatial reasoning
- Member of Artificial Social Intelligence group
Research & Publications
Learning Concept-based Causal Transition and Symbolic Reasoning for Visual Planning
- Yilue Qian, Peiyu Yu, Ying Nian Wu, Yao Su, Wei Wang, Lifeng Fan
- Published at IROS 2024 as Oral Pitch
- [Paper] [Web] [Video] [Poster]
Projects
Efficient Linear Algebra Solver with GPU Acceleration
Oct 2024 - Nov 2024
- Keywords: Cuda, Parallel Computation
- Developed a high-performance Linear Algebra Solver using C++ and Cuda, with advanced parallel techniques such as blocking, tiling, shared memory, and data shuffling within threads
- Achieved 82% of the performance efficiency of NVIDIA cutlass
- Conducted comprehensive benchmarking and profiling with NVIDIA Nsight
Efficient Compilation and Scheduling for CNNs on Compute-in-memory Units
Feb 2024 - Mar 2024
- Keywords: CNN, Compilation, Compute-in-memory
- Designed a highly efficient compilation approach for CNN inference on compute-in-memory chips
- Achieved 98% reduced memory access and 67% faster inference speed
- Implemented network reconstruction, memory access management, and computational operation scheduling
Evaluating and Mitigating Human-like Cognitive Biases in LLMs
Feb 2024 - Mar 2024
- Keywords: Large Language Models, Cognitive Bias, Benchmarking, Cognitive Hierarchy
- Established CoBAL, a comprehensive cognitive bias evaluation system
- Evaluated 7 cutting-edge language models (GPT-4, Gemini-1.5, Claude-3, etc.)
- Proposed HiTHer, a novel cognitive bias mitigating method with 40-85% higher performance
Playing Brick Breaker with A Robotic Arm System
Feb 2023 - Mar 2023
- Keywords: Robotic arms, Game AI, Inverse Kinematics
- Developed trajectory planning and inverse kinematics solutions
- Created a C++ library for inverse kinematics with 82% acceleration
- Implemented Pygame visualization with 60 fps performance
Playing Hanabi with ToM and Intrinsic Rewards
Sep 2022 - Dec 2022
- Keywords: Reinforcement Learning, Theory of Mind
- Designed goal-oriented intrinsic reward module improving game scores
- Implemented hand card inference module based on Theory of Mind
- Enhanced performance by 0.47 and 2.62 points in different game modes