Resume – Aixo

Leonard Li (Shang’ao Li) | 📞+86-189-7592-1544 | ✉️ leonardlsa@163.com

Education

Nanjing University (NJU), School of Computer Science B.Sc. in Information and Computing Science 09/2022 – 07/2026

GPA: 4.60/5.00 (Rank: 3/25)
Selected Courses: Machine Learning (95), Principles & Applications of LLMs (92), Knowledge Representation (96), Formal Languages & Automata (94), Operating Systems (91).
TOEFL: 101 (R29 / L26 / S23 / W23)

Ludwig Maximilian University of Munich · Research Intern · Supervisor: Dr. Yao Zhang
11/2025 – Present

Conducting research on advanced reasoning-based WebAgents, focusing on fine-grained reinforcement learning strategies.
Designing novel reward modeling mechanisms to enhance training stability and efficiency in RL-based WebAgent systems.
Exploring image-grounded reasoning and multimodal integration to improve WebAgent perception and decision-making.

KRistal Research Group, Nanjing University · Research Intern · Supervisor: Prof. YiZheng Zhao
10/2025 – Present

Developing a differentiable neuro-symbolic reasoning framework inspired by emALC semantics, aiming to enable learnable and adaptive inference processes.
Investigating meta-learning–driven reasoning strategies and constructing a Differentiable Fuzzy GNFO Engine that optimizes both logical fidelity and computational efficiency.

Independent Research · Research Intern · Supervisor: Dr. Zhen Han
07/2025 – Present

Researching reasoning-enhanced reward models for LLMs.
Designed pipelines combining reject sampling, SFT, and RL for scalable preference alignment.
Leading experiments on reasoning-guided reward modeling.
Manuscript in preparation.

ScaleML Lab, University of Illinois Urbana-Champaign · Research Intern · Supervisor: Prof. Tong Zhang
04/2025 – 06/2025

Integrated formal reasoning (Lean4 + mathlib) with LLMs to solve mathematical reasoning benchmarks (MiniF2F).
Proposed an interactive framework combining LLM guidance with Lean4 state-based verification.
Developed prototype enabling bidirectional communication between LLMs and Lean4, now extended by collaborators.
Co-first author manuscript in preparation.

Huawei Central Software Institute, 2012 Labs · Research Intern · Supervisor: JianFeng Gui
07/2025 – 09/2025

Independently co-developed a benchmarking system from scratch to evaluate the reasoning and adaptability of HarmonyOS intelligent agents.
Contributed to the IntelliOS-agent pipeline and participated in porting its Python dependency libraries to HarmonyOS.
Integrated HDC debugging pipelines with LLM-based reasoning modules, enhancing system stability and cross-platform compatibility.
System deployed in Huawei’s internal IntelliOS project.

Quantum Engineering & Simulation Theory (QUEST) Lab, NC State University · Research Intern · Supervisor: Prof. Yuan Liu
07/2024 – 11/2024

COSEC Research Group, Nanjing University · Research Assistant · Supervisors: Prof. Yuan Zhang, Prof. Sheng Zhong
07/2023 – 12/2024

Researched adversarial backdoors in ML models.
Proposed a novel exploit mechanism; designed attack experiments on malicious training pipelines.
Work contributed to group’s broader research on ML robustness & trustworthiness.

Mobile Robot Pose Estimation (AI and Robotics for Mobile Robot Manipulation – Boston Dynamics Project): Built a pose-estimation module with multi-method point-cloud denoising, improving robustness for mobile manipulation tasks.
Object Detection & Classification (AI in Visual Computing MIT Blended PBL): Implemented YOLO/Faster R-CNN pipelines and explored transfer learning for high-accuracy detection under limited data.
Gaussian SPH Fluid Simulation (Computer 3D Graphics and Deep Learning – NVIDIA Project): Developed a 3D Gaussian–SPH fluid pipeline integrating DFSPH constraints with anisotropic Gaussian rendering.
Autograd & Neural Network in C++: Reimplemented PyTorch’s autograd and constructed a fully connected neural net from scratch to study optimization and computation graphs.
Operating System Development (NJUOS & MIT xv6 labs): Implemented scheduling, memory management, and file systems; built a teaching-oriented OS prototype.
Retrieval-Augmented Generation (RAG) System: Built a full-stack Q&A/summarization system (fine-tuned LLaMA + vector DB + Vue/Spring Boot frontend).

National Scholarship, Ministry of Education of China — Highest national honor recognizing top 0.2% of students for academic excellence (2025)
First Prize (National Champion), HITCTF 2025 (Antiy Cup) — Ranked 1st nationwide in cybersecurity competition (2025)
First Prize, Entropy Cup Cryptography Challenge (CACR) (2025)
National Third Prize, China Graduate Information Security Innovation Competition (2024)
Special Scholarship for Fundamental Subjects, First Prize (twice) (2022–2024)
Outstanding Student Model, Nanjing University (2023–2024)
Provincial Award, China College Student Computer Design Competition (2023)

Teaching Assistant, Introduction to Information and Computing Science, Nanjing University (Prof. Sheng Zhong) · 2023–2024
Teaching Assistant, Discrete Mathematics, Nanjing University (Prof. Sheng Zhong) · 2023–2024
Teaching Assistant, Cryptography, Nanjing University (Prof. Yuan Zhang) · 2024–2025

Assisted in homework and exam grading, exam preparation, held weekly exercise classes, and provided student support.

Programming: C++, Python (primary), Rust, Java, Lua; familiar with Lean4.
Frameworks & Tools: PyTorch, HuggingFace, FSDP, DeepSpeed, Lean4/mathlib.
Systems: Linux (Arch, Fedora); HPC cluster & distributed training environments.