Vongani
Maluleke

I develop models that understand and generate human motion and social interaction, aiming to enable socially intelligent behavior in humanoids and autonomous systems. My research explores generative modeling and representation learning for multi-agent motion and human–humanoid interaction, spanning 3D vision, generative models, and AI for social good.

I am a Ph.D. student at UC Berkeley in BAIR, co-advised by Prof. Jitendra Malik and Prof. Angjoo Kanazawa. Before Berkeley, I worked as an AI consultant at Deloitte, delivering AI solutions across industries including logistics, banking, entertainment, insurance, and telecommunications.

  News BAIR Seminar: PhD Defence Talk — watch the recording.  On the Job Market — seeking research scientist & industry positions. 
Email Scholar Twitter LinkedIn GitHub
Vongani H. Maluleke
UC Berkeley
BAIR
Berkeley Artificial Intelligence Research
BAIR Seminar: PhD Defence Talk

Modeling Social Intelligence through Human Motion

BAIR · UC Berkeley May 6, 2026

Social intelligence (SI) unfolds largely through non-verbal signals such as timing, posture, and synchrony, enabling people to interact fluidly. By modeling these signals as motion in joint space, we can study how agents coordinate, anticipate, and adapt to one another in virtual and physical environments. Grounding SI in human motion provides a pathway toward systems that can perceive, understand, and participate in social interactions.

Read full abstract

In this talk, I present a line of work that explores this idea. Synergy and Synchrony in Couple Dances examines the extent to which one partner’s motion informs the other’s, using swing dance as a testbed for tightly coupled dyadic interaction. Diffusion Forcing for Multi-Agent Interaction Sequence Modeling extends this to a unified framework that supports partner prediction, inpainting, agentic generation, and joint future prediction within a single model that scales naturally with the number of agents. These ideas come together in ArcHumanoid, which brings SI into embodied systems, enabling humanoids to anticipate, react to, and coordinate with human partners in natural settings.

Also on YouTube Dissertation talk · BAIR

01 Publications

Under Review
Diffusion Forcing for Multi-Agent Interaction Sequence Modeling
Vongani H. Maluleke, Kie Horiuchi, Lea Wilken, Evonne Ng, Jitendra Malik, Angjoo Kanazawa
Project · arXiv · BibTeX · Blog
Under Review
Synergy and Synchrony in Couple Dances
Vongani H. Maluleke, Lea Müller, Jathushan Rajasegaran, Georgios Pavlakos, Shiry Ginosar, Angjoo Kanazawa, Jitendra Malik
Project · arXiv · BibTeX · Blog
Studying Bias in GANs through the Lens of Race
Vongani H. Maluleke, Neerja Thakkar, Tim Brooks, Ethan Weber, Trevor Darrell, Alexei A. Efros, Angjoo Kanazawa, Devin Guillory
Project · arXiv · BibTeX · Blog

02 Experience

  Education

Ph.D., Computer Science
UC Berkeley
3D Vision, Generative Models, AI for Social Good. Advised by Prof. Malik & Prof. Kanazawa.
2021–Present
MSc Advanced Analytics (Distinction)
University of Cape Town
Estimating poverty from aerial imagery with CNNs. Advised by Prof. Er & Dr. Williams.
2019
BBusSci in Analytics
University of Cape Town
Cryptic frog species identification from acoustic data. Advised by Prof. Durbach.
2016

  Industry

AI Senior Consultant
Deloitte Consulting, Johannesburg
2020–21
AI Consultant
Deloitte Consulting, Johannesburg
2019–20
Masters Studentship Candidate
CSIR, Pretoria
2017–19
Full Resume