Building agents that
reason, perceive, and act.
I am Yuhan Shu, an undergraduate entering the Agent / World Model / Embodied Intelligence space — using graphics and quantitative projects to prove I can turn technical ideas into rigorous working systems.
Public leaderboard · Agentic Alphas team
Advised by LAMDA faculty · reasoning & world models
Selected participant · 2026.08
Second author · Python-based physics study
From industry questions
to working products.
During my internship at Jiangsu Runlan Times New Energy Technology Co., Ltd., I contributed to two new-energy projects spanning industry research, product design, data engineering, and AI-enabled workflows. The company is the internship host; the Nanjing University Business School “Mantianxing” summer practice program is the project background.
New-Energy
Strategic Insight Platform
Built around a reviewable information workflow for the new-energy aftermarket: source collection, candidate review, controlled publishing, search, regional views, and team value evaluation.
NJU New-Energy
Industry Community
Local demo for a new-energy and new-materials industry community, covering access, profiles, demands, member search, messaging, and governance flows.
From ideas
to working systems.
My work sits at the intersection of intelligent decision-making and the environments in which decisions happen.
RealityScan
UV-to-3D Navigation
Built and validated a localization-to-navigation pipeline: 500/500 images registered, 25/25 paths completed, and collision violations held at zero in the released evaluation.
MathAgent
Demo
Solver–verifier workflow with local symbolic tools, batch experiments, and offline reports for reproducible mathematical reasoning.
Quant Factor
Research Lab
Modular factor research framework with 17 categories, 131 definitions, IC/ICIR evaluation, walk-forward validation, FDR control, and cost-aware portfolios.
Synera
Auto Arena
Object-oriented auto-battler prototype with state-machine logic, simulation-ready rules, and a Qt/QML interface.
Reading deeply.
Building with intent.
Currently reading and planning research around agents, world models, embodied intelligence, and graphics — with quantitative modeling as a parallel technical practice.
See the code ↗