University of Verona (UNIVR)
Studying the generalization of activity recognition models on a multi-centric dataset. [Tech Stack: Python, PyTorch, OpenCV]
October 2022 - Present
Topic: Multi-level surgical activity recognition from endoscopic videos
University of Verona (UNIVR) & University of Strasbourg (UNISTRA)
Developed deep learning model for surgical activity recognition from endoscopic videos at two levels of granularity: Phase & Step. [Tech Stack: Python, TensorFlow, PyTorch, OpenCV]
Researched weakly supervised learning method for finer-grained activity recognition (step) using weak coarser (phase) labels. [Tech Stack: Python, PyTorch]
Designed and developed a simplified temporal augmentation method for effective training of spatio-temporal models for the task of surgical activity recognition. [Tech Stack: Python, PyTorch]
Led a collaborative study on the effectiveness of self-supervised learning methods in the context of surgical computer vision. [Tech Stack: Python, PyTorch, VISSL]
Worked on cleaning, structuring, and analyzing surgical video datasets. Collaborated with clinicians to understand surgical workflows, annotation protocols, and manage surgical datasets. [Tech Stack: Python, OpenCV, JSON]
October 2019 - September 2022
Ph.D. Topic: Multi-level surgical activity recognition from endoscopic videos
Advisors:
Prof. Paolo Fiorini, UNIVR
Prof. Nicolas Padoy, UNISTRA
Aug 2017 - September 2019
Developed LiDAR- and camera-based SLAM algorithms for accurate localization of the autonomous vehicle of Ati Motors in manufacturing plants. The LiDAR-based SLAM algorithm has been a key component of the vehicle’s autonomy stack contributing to the success of the company. [Tech Stack: Python, C++, PCL]
Implemented an urban traffic simulator that aided in the research and development of trajectory planning and control algorithms for the vehicle. [Tech Stack: Unity, C#]
Investigated reinforcement learning techniques, such as deep Q-learning and policy gradient method, to learn optimal control and driving policies of the vehicle. [Tech Stack: Python, Tensorflow, OpenAI Gym]
Collected, organized, and analyzed multi-modal dataset (LiDAR, Stereo Cameras, IMU) for SLAM, Object detection and tracking, lane detection, etc. [Tech Stack: Python, C++, PCL, PyTorch, Protobuf]
International Institute of Information Technology, Bangalore (IIITB)
August 2012 - July 2017
May 2016 - July 2016
Developed an intelligent chatbot for handling all company’s products and services related text/image queries. [Tech Stack: Python, OpenCV]
Responsibilities included building a prototype of a recommendation engine for products suggestion.
May 2015 - July 2015
Worked on the scalability and testing of one of their binary instrumentation tool. [Tech Stack: C++, IntelPin]
Responsibilities included performance evaluation, code analysis, and identifying suitable code coverage tools.