LightSim
Traffic simulation for San Francisco’s Financial District, built on the Cell Transmission Model.
Open source · Traffic systemsI’m Brooklyn (Haoran Su), a Staff ML Engineer at TikTok, focused on large recommendation models and inference engineering. My work spans retrieval and ranking, production deployment, and model performance optimization.
My research explores deep transformers, multi-agent learning, and intelligent traffic systems. I hold a Ph.D. in Engineering from NYU and a B.Sc., B.A., and M.Sc. from UC Berkeley.
Selected papers
Multi-Agent Decision Transformer
Spatiotemporal sequence modeling for multi-agent control.
Manifold-Geometric Transformer
Geometric and dynamic scaling in deep transformers.
Hierarchical GNN Corridor Formation
Multi-agent reinforcement learning for emergency vehicle corridors.
EMVLight
Coordinating emergency vehicle routing and traffic signal control.
Transportation Research Part C