I am working as a Research Engineer with the Microsoft AirSim team.
Before that, I was a Masters in Robotics student at Carnegie Mellon University, working on perception and motion planning algorithms for quadrotors at the Air Lab at the Robotics Institute, advised by Sebastian Scherer and working closely with Michael Kaess. I focused on thin obstacle detection and mapping for UAVs, and marrying traditional sampling based motion planning algorithms with machine learning.
During times when I am tired of being a wannabe in robotics, I tend to come back to old hobbies or try to pick up some new ones. These include biking, reading (totally not adding books to my to-read list), playing the guitar pretentiously, sketching, and writing.
(Not Fake) News
- 2020-03: A pre-print for AirSim Drone Racing Lab is now available. We used our framework in the Game of Drones competition at NeurIPS 2019.
- 2020-03: An updated version of our pre-print, Learning Visuomotor Policies for Aerial Navigation Using Cross-Modal Representations is now available. See this blog post for an overview.
- 2019-12: I helped organize Game of Drones, a drone racing competition at NeurIPS 2019 built on AirSim. Read our MSR blog post for a quick overview.
- 2019-09: Check out our pre-print and video on learning cross-modal representations for imitation learning, applied to the drone racing problem. Work done with our summer research intern, Rogerio Bonatti.
- 2019-01: Multi-view Reconstruction of Wires using a Catenary Model was accepted at ICRA 2019!
- 2018-11: Moved to Seattle and started working at Microsoft. The winter days seem even shorter and darker than Pittsburgh, but at least, there is no snow. Pittsburgh robotics vibe is unmatched, but the Pacific Northwest has its own charm to it.
- 2018-10: Learning Adaptive Sampling Distributions for Motion Planning by Self-Imitation was accepted to the Machine Learning in Robot Motion Planning workshop in IROS 2018.
- 2018-08: Defended my Masters Thesis, “Wire Detection, Reconstruction, and Avoidance for Unmanned Aerial Vehicles”, spanning semantic segmentation using synthetic data generation and dilated convolutional networks, model-based multiview reconstruction, model-free grid mapping, and trajectory library based avoidance.
For more details, here is the defense video, thesis slides, and the PDF.
- 2018-06: DROAN - Disparity-space Representation for Obstacle AvoidaNce : Enabling Wire Mapping & Avoidance was accepted to IROS 2018.
- 2017-07: Wire Detection using Synthetic Data and Dilated Convolutional Networks for Unmanned Aerial Vehicles was selected as a finalist for the Best Application Paper award at IROS 2017.
- 2017-07: Deep Flight: Learning Reactive Policies for Quadrotor Navigation with Deep Reinforcement Learning appeared in the Workshop on Learning Perception and Control for Autonomous Flight at RSS 2017.
- 2016-08: Started Masters in Robotics (Research) at Carnegie Mellon University, Pittsburgh.
- 2016-05: Finished RA-ship at the AirLab, RI, CMU. I was an RA from 2015-09 to 2015-11, and then from 2016-02 to 2016-05.
- 2015-08: Finished the best summer internship at the Manipulation Lab, RI, CMU as part of the 2015 RI Summer Scholars cohort. I worked on predicting resulting orientation of cubes when dropped from random heights using Bingham distributions. See report and poster.
- 2015-05: Finished undergrad at Indian Institute of Technology, Roorkee. After 2 weeks, flew to Pittsburgh to begin a new life.