Bo Chen

Senior Staff Machine Learning Scientist

Seeing Machines — Core AI

Biography

I am a Senior Staff Machine Learning Scientist in the Core AI team at Seeing Machines, specializing in human-centered 3D perception for Physical AI, robotics, and safety-critical systems.

I lead the development of real-time Human Mesh Recovery (HMR) and human-understanding technologies, combining 3D geometry, human body modelling, multimodal learning, and efficient on-device inference. I enjoy setting technical direction, developing novel ML architectures, and translating research into real-time embedded systems, integrated perception platforms, and high-profile product demonstrations — most recently the Interior Perception Map showcased at CES 2026.

Previously, I was a Research Fellow at the Australian Institute for Machine Learning, The University of Adelaide, working on 6D object pose estimation, differentiable geometric optimization, adversarial machine learning, and efficient edge AI. Our team won two back-to-back first-place outcomes in the European Space Agency’s global satellite pose estimation competitions.

I also serve as a peer reviewer for IEEE TPAMI, CVPR, ICCV, AAAI, ACM Multimedia, IJCAI, ICRA, ICDM, WACV, and IEEE Access.

Interests

Human-Centered 3D Perception
Human Mesh Recovery
Physical AI & Robotics
3D Vision & Geometry
Efficient On-Device AI

Education

PhD in Artificial Intelligence, 2018

Monash University
MSc in Statistics, 2014

The University of Queensland
Bachelor of Management in E-Commerce, 2006

Sun Yat-Sen University

Experience

Senior Staff Machine Learning Scientist

Seeing Machines — Core AI

Feb 2022 – Present Canberra, Australia

Human Mesh Recovery for Physical AI: Led the development of a real-time HMR system for Physical AI and robotics, driving model architecture, technical direction, and performance optimization. Developed geometry-aware and semantics-driven architectures that outperformed Meta’s then-state-of-the-art SAM3D-Body in accuracy while achieving 40× lower inference latency, with ~15 ms embedded inference on NVIDIA Jetson Thor.
Human understanding for robotics: Extended human-understanding technology beyond in-cabin perception toward Physical AI applications, including human-aware navigation, teleoperation, and robot-environment interaction. Authored a strategic whitepaper defining the technical direction for Seeing Machines' human-centered AI capabilities.
Interior Perception Map (IPM): Led the human-understanding component of Seeing Machines' next-generation unified perception architecture demonstrated at CES 2026, introducing multi-view occupant reconstruction for detailed 3D human modelling within a holistic cabin digital twin.
End-to-end system delivery: Key technical contributions across perception architecture, model development, system integration, synthetic-data-driven development, and real-time optimization.
Broader Core AI: Led and contributed to initiatives spanning robust 3D perception, multimodal-LLM temporal grounding, video anomaly detection using normalizing flows, synthetic-data generation, and LLM bias calibration.

Research Fellow

Australian Institute for Machine Learning (AIML), The University of Adelaide

Feb 2019 – Feb 2022 Adelaide, Australia

3D vision and machine learning research: Developed novel methods in 6D object pose estimation, differentiable geometric optimization, adversarial machine learning, and efficient edge AI.
Competitive research: Achieved two back-to-back first-place outcomes in global AI competitions organized by the European Space Agency (ESA) for satellite pose estimation.
Research leadership and mentorship: Supervised and mentored PhD and MPhil students across problem formulation, algorithm development, experimentation, evaluation, and publication.

Teaching Associate

Monash University

Jul 2017 – Dec 2018 Melbourne, Australia

Taught postgraduate data science and machine learning, covering Python, R, statistical modelling, classical machine learning, and neural networks.

Data Analysis Lead

Moonbasa

Jul 2011 – Jun 2012 Guangzhou, China

Assistant Research Executive

Ipsos

Sep 2010 – Feb 2011 Guangzhou, China

Awards

Pose Estimation 2021 AI Competition — 1st and 3rd place

European Space Agency (ESA) 2022

1st and 3rd place across the two dataset categories of the second Satellite Pose Estimation Competition (SPEC2021).

Satellite Pose Estimation Challenge — First Place

European Space Agency (ESA) Kelvins 2019

Won first place in this global AI competition held by the European Space Agency.

See certificate

Featured Publications

Bo Chen, Tat-Jun Chin, Marius Klimavicius

October 2021 In WACV 2022

Occlusion-Robust Object Pose Estimation with Holistic Representation

We propose novel techniques to make object pose estimation remarkably robust against occlusions.

Bo Chen, Alvaro Parra, Jiewei Cao, Nan Li, Tat-Jun Chin

June 2020 In CVPR 2020

End-to-End Learnable Geometric Vision by Backpropagating PnP Optimization

We propose BPnP, a network module for backpropagation through a PnP optimizer, as if the optimizer were a differentiable function.

Bo Chen, Jiewei Cao, Alvaro Parra, Tat-Jun Chin

September 2019 In ICCVW 2019

Satellite Pose Estimation with Deep Landmark Regression and Nonlinear Pose Refinement

We proposed an approach to estimate the 6DOF pose of a satellite. Our method won the first place in the Kelvins Pose Estimation Challenge organised by the European Space Agency (ESA).