Bin Zhang

Ph.D., Senior Researcher in Autonomous Driving | Ex. Staff Engr. Baidu Apollo

Shanghai, China

About

I am passionate about accelerating the scaling laws of autonomous driving and building safer, more intelligent driving systems through AI innovation. I am particularly interested in scalable end-to-end autonomous driving systems, uncertainty-aware learning, and the next generation of foundation models. Previously, I was a Staff Engineer at Baidu Apollo, where I designed a game-theoretic POMDP planner that reduced lane-merge-related MPCI by 90%. I also led the "Straight Driving Scene" team to shift from rule-based to learning-centric planning, which contributed to a 70% reduction in accident rate. During my Ph.D., I focused on the representation and propagation of general-form uncertainty in neural networks, and the methodology was applied to projects with our partners, including GE and Boeing, bridging theoretical development with real-world impact.

Experience

  • Senior Researcher at Shanghai Qi Zhi Institute
    May 2025 - Present · 1 yr 3 mos

    Recipient of the Shanghai Magnolia Pujiang Talent Program Funding Research on scalable foundation models for autonomous driving—spanning VLA, world models, and RL—actively seeking academic and industry partnerships to accelerate real-world deployment.

  • Staff Engineer, Decision & Planning for L4 Autonomous Driving at Baidu Apollo
    Dec 2021 - May 2025 · 3 yrs 6 mos

    Designed a game-theoretic POMDP framework for ego planning in dynamical lane merge scenarios, proposed a dual-Bayesian scheme to assess the uncertain intentions of other vehicles in the target traffic flow and reduced the merge-related MPCI by 90%. Led the "Straight Driving Scene" team within the Planning & Control department, which accounts for 70%+ of mileage. Driven the architecture upgrade from rule-based to learning-centric planning and reduced our team's accident rate by 70%.

  • Graduate Research And Teaching Assistant at Purdue University
    Aug 2015 - Aug 2021 · 6 yrs 1 mo

    Conducted research on data-driven science: neural networks, probabilistic uncertainty analysis, Bayesian machine learning, and fuzzy system, with accomplished projects for industrial partners (GE, Boeing, etc.) that solved production problems using ML/CV methods. Highly skilled in state estimator (Kalman and particle filters) and controller (PID, LQR, MPC, T-S fuzzy, etc.) designs; Expertise in signal/image processing, and Python/C++ programming.

  • Visiting Scholar at The University of British Columbia
    Aug 2013 - Oct 2013 · 3 mos

    Working on physics-based optimization of milling operations under the direction of Prof. Yusuf Altintas.