Computer Vision · Robotics
I am a Computer Vision Engineer (R&D) at Ogmen Robotics, where I research and build perception models for understanding pet and animal behavior.
My broader research goal is to develop robot learning systems that learn from visual and tactile perception without relying on dense supervision or large-scale datasets.
I am particularly interested in the low-data regime, where robotic systems must adapt across diverse environments and generalize to novel tasks with minimal prior exposure. This includes learning sample-efficient policies that can robustly transfer across tasks and real-world settings.
More broadly, I aim to build autonomous robotic systems capable of performing complex, sensitive, and dexterous human-centric tasks.
I completed my undergraduate degree in Robotics and Automation from Symbiosis Institute of Technology, Pune, in 2025, where I was advised by late Prof. Javed Sayyad. Prior to joining Ogmen Robotics full-time, I interned with the company for one year.
* denotes equal contribution.
K9-Bench: Evaluating Multimodal LLMs on Canine-Centric Videos
arXiv Preprint, 2026
K9-Bench is a benchmark of 4,744 QA pairs across 907 canine videos for evaluating MLLMs, highlighting strong zero-shot performance while exposing limitations in canine activity understanding and revealing gaps in compositional and temporal reasoning.
GEPAF: A Non-Monotonic Generalized Activation Function in Neural Network for Improving Prediction with Diverse Data Distribution Characteristics
Neural Networks, Vol. 180, December 2024
A non-monotonic generalized activation function that improves neural network prediction across diverse and complex data distributions.
I grew up in Mumbai and studied in Pune. Outside of research, I enjoy playing Tabla — I have cleared a few practical examinations in the music.
I am open to research conversations, collaborations, and new opportunities. The best way to reach me is by email.
Services: Reviewer for IJCNN 2025/26/27; Neural Networks and many more.