I am a Machine Learning Researcher at Qualcomm AI Research. My research interests lie in the areas of Large Language Models, Generative Models and Bayesian Inference. Previously, I was a Postdoc at the University of Tuebingen working in the Autonomous Vision Group. Before that I was a PhD student at the Max Planck Institut für Informatik, Saarbrücken, advised by Dr. Bernt Schiele and Dr. Mario Fritz (My PhD thesis is available here). I completed my Master’s Thesis at Saarland University under the supervision of Dr. Jilles Vreeken in the area of Algorithmic Data Mining and my Bachelors degree at the National Institute of Technology, Karnataka, India.

News

• Our work “Look, Remember and Reason” will appear at the KLR workshop at ICML 2023.
• Check out our work on using LLMs as fitness coaches at the Embodied AI Workshop at CVPR 2023, video and blogpost.
• Check out our sdfstudio repo: A Unified Framework for Surface Reconstruction.

Publications

Look, Remember and Reason: Visual Reasoning with Grounded Rationales,
Technical Report, 2023.
Paper


KING: Generating Safety-Critical Driving Scenarios for
Robust Imitation via Kinematics Gradients,
European Conference on Computer Vision (ECCV), 2022 (oral).
Paper | Project Page


Euro-PVI: Pedestrian Vehicle Interactions in Dense Urban Centers,
Computer Vision and Pattern Recognition (CVPR), 2021.
Paper | Data | Code


Normalizing Flows with Multi-scale Autoregressive Priors,
Computer Vision and Pattern Recognition (CVPR), 2020.
Paper | Workshop Extension | Code | Video


Haar Wavelet based Block Autoregressive Flows for Trajectories,
German Conference on Pattern Recognition, 2020 (oral).
Paper | Video


Updates-Leak: Data Set Inference and Reconstruction Attacks in Online Learning,
USENIX Security Symposium (USENIX Security), 2020.
Paper


Conditional Flow Variational Autoencoders for Structured Sequence Prediction,
BDL@NeurIPS’19 and ML4AD@NeurIPS’19 (oral).
Paper | Video


“Best-of-Many-Samples” Distribution Matching,
BDL@NeurIPS’19.
Paper


Bayesian Prediction of Future Street Scenes using Synthetic Likelihoods,
International Conference on Learning Representations (ICLR), 2019.
Paper | Code + Data


Accurate and Diverse Sampling of Sequences based on a “Best of Many” Sample Objective,
Computer Vision and Pattern Recognition (CVPR), 2018 (oral).
Paper | Code


Long-Term On-Board Prediction of People in Traffic Scenes under Uncertainty,
Computer Vision and Pattern Recognition (CVPR), 2018.
Paper | Code + Data


Long-Term Image Boundary Prediction,
AAAI Conference on Artificial Intelligence, 2018.
Paper


Efficiently Summarising Event Sequences with Rich Interleaving Patterns,
SIAM International Conference on Data Mining (SDM), 2017.
Paper