Dec 2024 – Jan 2026 · 1 yr 1 mo · London Area, United Kingdom · On-site

Aug 2022 – Aug 2023 · 1 yr · New York, United States · On-site

Sep 2019 – Aug 2020 · 11 mos · Montreal, Canada Area

• Working on Probabilistic Deep Learning and Uncertainty Estimation

Aug 2018 – Apr 2019 · 8 mos · Kharagpur Area, India

• Modeled driving as multi-agent RL problem employing suitable reward, state and action space.
• Every agent used actor-critic network for deciding policy, along with Ornstein-Uhlenbeck method for exploration

May 2018 – Jul 2018 · 2 mos · Gurgaon, India

• Developed a real-time image dehazing/defogging method for autonomous vehicles for low end hardware
• Achieved 10X speed up than prior arts, model runs at 200 FPS & 20 FPS on NVIDIA 1080 ti and Jetson TX1
• Increased the object detection accuracy by about 65% in presence of dense fog using the designed GAN model
• Implemented 4 state-of-art image dehazing papers in tensorflow, creating open-source dehazing model zoo
• Created a synthetic foggy image dataset for Indian road scenes from clear images and depth maps

May 2017 – Aug 2017 · 3 mos · Kharagpur Area, India

• · Developed a deep learning based Visual Odometry pipeline to find egomotion of a robot using successive images from monocular camera
• · Convolutional and Recurrent Neural Network are the main component of model, used in series, taking spatial and temporal features in account and regression loss is used to predict lateral and rotational displacement of bot between image frames

Jan 2017 – Mar 2017 · 2 mos · Kharagpur Area, India

• · Developed a large scale robot to carry disabled people up and down stairs while also being able to move on floors and capable to autonomously navigate through the mapped rooms avoiding obstacles
• · Developed vision based autonomous navigation package having real time object-tracking and obstacle avoidance
• · Developed Embedded Design of the bot, included control of actuators, brain based signal control, voice control and touch interface

Feb 2016 – Apr 2019 · 3 yrs 2 mos · IIT Kharagpur

• ·Worked on solving Autonomous Vehicle Navigation challenges such as obstacle avoidance, road segmentation and traffic light detection using deep learning based models such as SegNet, YOLO, RCNN.
• · Developed the ROS based embedded system and control system for an Autonomous Mobile Robot, Eklavya 5.0 which participated in Intelligent Ground Vehicle competition 2016 held at Oakland University, Michigan, USA
• · Building a self-driving car for Mahindra Rise Driverless Car Challenge, our team was selected among top 13 teams out of the 600+ and awarded a Electric Car Mahindra E2O
• · Worked on sensor integration, designing control system and developed CAN bus interface of Mahindar E2O for communication between user, sensors and car inbuilt actuators

Sep 2015 – Feb 2016 · 5 mos

• I worked on making front end of website for Global Entrepreneurship Summit 2017 by E Cell, IIT Kharagpur. Mainly worked with HTML, CSS, Bootstrap, Java script and PHP.