Hello! 👋

Bonjour! 👋

Hola! 👋

Zdravstvuyte! 👋

Nǐn hǎo! 👋

Marhabaan! 👋

Kon'nichiwa! 👋

Hallo! 👋

Ciao! 👋

Olá! 👋

Hallo! 👋

Cześć! 👋

Hallå! 👋

Xin chào! 👋

Halo! 👋

Yassas! 👋

Annyeonghaseyo! 👋

Bună! 👋

Helló! 👋

Ahoj! 👋

Experience

Research Assisstant
Machine Learning Lab, IIIT Hyderabad
2023 September - Present

Research Intern
Robotics Research Center, IIIT Hyderabad
2023 January - 2023 September

A pipeline was created to generate detailed top-down maps enriched with instance data, significantly improving the Visual Language Navigation of an autonomous wheelchair. Additionally, a gesture-controlled robot was engineered to follow a person while avoiding static obstacles, enhancing its navigation and interaction capabilities. These advancements have promising applications in fields like healthcare and logistics.

Co-Founder
Flowdrive.ai
2021 April - Present

I helped develop an open-source, cross-platform Level 2 Advanced Driver Assistance System (ADAS) by using my expertise in robotics and artificial intelligence. We made sure it worked with various hardware configurations and improved its reliability and performance through careful data management. We also tested it extensively, especially in real-world scenarios using the CARLA simulator framework. These efforts resulted in a strong and highly functional ADAS system.

Data Science Fellow
Fellowship.ai
2022 January - 2022 March

During my tenure, I engaged in rigorous Research and Development (R&D) efforts focused on video classification models, pose estimation techniques, and video analytics. My contributions were instrumental in shaping algorithms and tools designed for the comprehensive analysis of video content. This encompassed pivotal aspects such as feature extraction, meticulous data cleaning, and insightful visualization techniques. My work in this domain was driven by a commitment to advancing the state-of-the-art in video analysis, ultimately contributing to enhanced insights and utility in this dynamic field.

Junior Machine Learning Engineer
Omdena AI
2021 June - 2021 August

A real-time object recognition system has been developed and deployed, focusing on the detection of buses and the recognition of their license plates using Optical Character Recognition (OCR) models. These models have been seamlessly integrated into a mobile application for real-time bus detection and seat occupancy monitoring. To ensure optimal performance, considerable effort has been dedicated to model optimization, aiming for efficient processing and low latency. This integrated solution significantly enhances bus tracking and occupancy management within the mobile app.

Core Member (ML Team)
Developer Student Club, TIET Patiala
2020 April - 2020 September

A pseudo LIDAR system has been developed using cameras and deep learning techniques, utilizing DenseDepth for 3D reconstruction from monocular images. Multiple deep learning models have also been developed and trained for behavioral cloning in self-driving cars, enabling them to emulate human driving behavior. In addition, lane detection algorithms have been integrated into these systems, providing robust lane tracking and navigation capabilities. This holistic approach combines 3D scene understanding, behavioral modeling, and lane detection to improve the performance and safety of autonomous vehicles.

Publications

Physical Reasoning and Object Planning for Household Embodied Agents

Transactions on Machine Learning Research, , , , 2024

Instance-Level Semantic Maps for Vision Language Navigation

Proceedings IEEE Conf. on Robot and Human Interactive Communication (RO-MAN) , 2023