About Me

Hello! I am Parnika, a research engineer focused on reinforcement learning, robot imitation learning, and the real-time deployment of learned models on robotic hardware. I hold an M.Sc. in Computer Science (Machine Learning) from the Université de Montréal and Mila — Quebec AI Institute.

At Mila’s REAL Lab, under the supervision of Prof. Glen Berseth, I designed a transformer-based actor-critic architecture with a novel tokenization scheme that trains a single reinforcement-learning policy across multiple robot morphologies. By conditioning the policy on BERT-encoded URDF (robot-structure) descriptions, it generalizes to robot embodiments unseen during training — without per-robot retraining (report). Earlier, I collaborated with Prof. Shalabh Bhatnagar and Prof. Raghuram Bharadwaj at the Indian Institute of Science on attention-based actor-critic methods for constrained cooperative multi-agent reinforcement learning, published at AAMAS 2021.

More recently, I set out to complement my simulation research with hands-on robot experience: I purchased an SO-101 robotic arm and built an end-to-end imitation-learning pipeline — from teleoperated data collection to on-hardware policy deployment — fine-tuning ACT, π0.5, and SmolVLA and evaluating them directly on the arm (results and failure analysis).

On the industry side, I worked as a Technologist at Wipro Ltd., where I optimized a semantic-segmentation model for real-time scene understanding on an autonomous-vehicle platform — converting it to an INT8-quantized TensorRT engine and achieving a 3× inference speedup on an NVIDIA Jetson ORIN. Before that, I was a Research Engineer at Mindtree Ltd., implementing research papers for customer use cases across reinforcement learning, NLP and computer vision. I completed my B.Tech. in Computer Science & Engineering at Dr. A.P.J. Abdul Kalam Technical University, India.

I am keen on growing as an end-to-end robotics engineer across learning, perception, and deployment. I work primarily in Python and PyTorch.

News

July 2026. Trained and evaluated π0.5 and SmolVLA imitation-learning policies on a custom SO-101 dataset — see Robot Learning.
June 2026. Deployed first imitation-learning policy (ACT) on a physical SO-101 arm.
June 2025. Completed Master’s at UdeM/Mila.
May–Dec 2024. Research internship at Mila’s REAL Lab, supervised by Prof. Glen Berseth. Report
November 2023. Awarded Scholarship from Ministry of Higher Education, Government of Quebec. C$7000
October 2023. Awarded University of Montreal Exemption Scholarship for international students, 2nd cycle. C$10000
September 2023. Started Master’s at UdeM/Mila.
September 2022. Technologist at Wipro Ltd.
May 2021. Presented poster at AAMAS.
December 2020. Our paper got accepted at AAMAS 2021.
October 2019. Started Research Collaboration with IISc, Bengaluru.
July 2019. Awarded Outstanding Performer, Mindtree.
November 2017. Started Internship and then joined as Research Engineer at Mindtree Ltd.