I'm an undergraduate student at Peking University, majoring in Intelligent Science and Technology (AI) and minoring in Economics.
I'm interested in AI Agents, real-world workflow automation, multi-agent collaboration, demand analysis, and product design. I'm currently working on an AI Agent project for automating financial and accounting workflows for small and medium-sized enterprises. At the same time, I'm learning more about Agent systems, Coding Agent techniques, multi-agent collaboration, and how to design valuable products based on real user needs.
Rather than simply chasing new concepts, I'm more interested in how AI can actually enter real-world scenarios: how to understand users' real problems, break down complex workflows, and coordinate models, tools, data, and humans to reliably complete tasks that used to require a large amount of manual work.
-
Financial & Accounting Workflow Automation AI Agent
- Exploring how AI Agents can process messy raw materials such as invoices, contracts, bank statements, and non-standard Excel files in SME financial scenarios
- Focusing on structured extraction, data cleaning, reconciliation, calculation, report generation, human review, and audit trails
-
AI Agents and Coding Agents
- Learning about task decomposition, tool use, context management, long-term memory, and workflow orchestration in Agent systems
- Exploring how to use Coding Agents more effectively for real engineering tasks
- Paying attention to tools and methods such as Claude Code, OpenClaw, and Skills
-
Multi-Agent Collaboration
- Interested in how multiple Agents can divide work, collaborate, communicate, and form stable workflows
- Thinking about the potential applications of multi-agent systems in software development, knowledge management, and enterprise workflow automation
-
Demand Analysis and Product Design
- Learning how to discover problems from real users and real scenarios
- Practicing user research, user interviews, feature prioritization, MVP design, and product iteration
P-Lib, formerly known as Pkuhub, is a campus resource-sharing platform for students at Peking University.
To me, this project is not just a technical project, but also a full experience in product building and organizational practice. I participated in multiple stages of the project, including user research, product design, resource cold start, and platform operations.
- Positioning: A platform for PKU students to share course materials and learning experiences
- Achievements: 4,000+ PKU users, 1,000+ shared materials, and 20,000+ downloads
- My work:
- Interviewed users from humanities, sciences, medical departments at PKU, as well as students from C9 universities
- Identified core needs around uploading, searching, credibility, and incentive mechanisms
- Contributed to product design and iteration
- Mobilized personal networks to collect the first batch of high-quality notes and materials
- Coordinated access to a 16c16g server to support development
- Helped complete the handover of the project to the next generation of students
P-Lib gave me my first systematic understanding that a platform is not finished once it is built. The real challenges lie in identifying genuine needs, completing the cold start, maintaining operations, and allowing the platform to grow naturally among users.
I'm currently working on an AI Agent project for automating financial and accounting workflows for small and medium-sized enterprises.
Many companies already have financial software, but before data can enter these systems, there is still a large amount of messy manual work: invoice photos, contract PDFs, bank statements, non-standard Excel files, and business ledgers need to be extracted, cleaned, reconciled, summarized, split, and then converted into reports or accounting vouchers.
I hope to build an intelligent preprocessing layer between raw business data and financial systems:
- Supports input from invoices, contracts, Excel files, bank statements, and other raw materials
- Automatically performs structured extraction, data cleaning, reconciliation, and calculation
- Generates reviewable Excel files, reports, or import-ready formats for financial systems
- Emphasizes auditability, traceability, and reproducibility
- Allows humans to review and correct results at key steps
This project is still under active iteration. I'm validating its demand boundaries, MVP priorities, and delivery methods through enterprise internship experience, industry interviews, and observation of real financial workflows.
- Python
- C++
- CUDA
- PyTorch
- Hugging Face Transformers
- Scikit-learn
- Pandas
- NumPy
- LLM application development
- Agent system basics
- Claude Code
- OpenClaw
- Skills
- Linux / Ubuntu
- Docker
- CMake
- Pixi
- User research
- Demand analysis
- Product design
- Workflow analysis
- Excel automation
- Financial data processing
- PKUCPC: Third Prize in 2024, Second Prize in 2025
- Chinese Chemistry Olympiad: Provincial First Prize
- Email: [email protected]
- Website: https://fomalhaut647.com
Beyond coding and building projects, I enjoy reading sociology-related books, learning all kinds of new knowledge, observing how complex systems operate, and thinking about how things may develop in the future.
I enjoy studying the underlying principles behind complex problems: why a system works the way it does, why an organization forms a certain structure, why a market shows a particular trend, and why a technology changes human behavior. I also enjoy forecasting future developments, learning from different fields, and experiencing new, interesting, and unfamiliar things.
To me, the most interesting part of AI is not only model capability itself, but also the way it gives us a chance to rethink information, decision-making, collaboration, automation, and how humans work together with tools to accomplish complex tasks.
Thanks for visiting! 😊