Back at AI Grid Summit in Berlin, discussing industrial AI, research impact, and the AI community. Read my reflections ↗
Researcher · Berlin, Germany
Vivek Chavan.
Machine learning, computer vision
& robotics.
I study how intelligent systems can learn from human activity, understand procedures, and act reliably in the physical world.
PhD candidate at TU Berlin
Research Associate at Fraunhofer IPK
Research focus
Understanding the world.
Learning to act in it.
01Multimodal ego/exo perception
02Procedural understanding & robot learning
03Data-centric & continual learning
Recent activity
A few updates
Presented recent research at ECCV in Malmö, including oral presentations at X-Reason and ACVR. Read my reflections ↗
Our work on multimodal worker guidance is published in CIRP Annals. Read the paper ↗
Featured in Fraunhofer IPK’s FUTUR on AI for the circular economy. Read the feature ↗
Selected work
Recent research
ECCV 2026 · DexHAND Workshop
What Matters, When? Diagnosing and Improving Conditional Visual Grounding in Visuomotor Imitation Policies
Understanding why robot policies select the wrong object or destination, and improving robustness to visual distractors.
Non-archival extended abstract. An extended manuscript is currently under review.
ECCV 2026 · X-Reason Workshop · Oral
Towards Neuro-Symbolic Procedural Reasoning for Long-Horizon Vision-Language-Action Manipulation
Combining VLA policies with task graphs and procedural memory for multi-step manipulation.
Non-archival extended abstract; public preprint.
NeurIPS 2025 · Datasets & Benchmarks
IndEgo: A Dataset of Industrial Scenarios and Collaborative Work for Egocentric Assistants
Multimodal ego/exo recordings of industrial work, supporting research on procedural understanding and egocentric assistants.
Beyond the papers
Notes from research
Get in touch
Research is a conversation.
I welcome discussions about research collaborations, speaking, and student projects.