• Developing adaptive optics with event cameras for navigation, dynamically adjusting focal length and aperture for depth estimation.• Implementing RL-based control for optimized navigation and obstacle avoidance.• Built a custom simulator using Gaussian splats to generate realistic, high frame-rate event data for HITL testing.

• Worked on real-time activity recognition vision stack using single fisheye images for indoor environments.• Investigated LLaVA, BCLIP, and CLIP for multi-modal vision-language tasks, leveraging LoRA for PEFT fine-tuning• Analyzed and deployed state-of-the-art Vision Transformers, enhancing object detection performance• Implemented CI/CD workflows using Jenkins and Kubernetes, automating the model lifecycle from data preprocessing to training, evaluation, and deployment in Dockerized environments

Developed hardware accelerated Computer Vision products addressing crucial real-world problems. Gate-Guard:• Developed a high accuracy edge-based ANPR (Automatic Number Plate Recognition) system.• Built optimized C++ pipelines using DeepStream-TensorRT to facilitate real-time actuations. • Increased mAP of existing OCR network by 20% by addressing colour-invariant plate recognition problem.• Successfully designed and implemented synchronous analytics and monitoring dashboard.

Led a team of 56 passionate undergrad engineers on multiple national and international Formula Student platforms.Performance of the team under my leadership:• 2019 Formula Bharat, Kari Motor Speedway, Tamil Nadu, India - Overall rank 5 out of 63 best national teams. - Winners of the Cost presentation event. - Finalist in the engineering design event. - Best Electric Powertrain Package in Electric Vehicle Concept Challenge.• 2019 Formula Student Germany, Hockenheimring, Hockenheim, Germany - World rank 9th among 150 qualified teams in the qualification quiz attempted by ~300 teams. - Fastest acceleration with 0-70kmph in 4.6 secs in the Combustion race acr discipline. - #1 Indian team defeating all IITs and other highly ranked college teams. - World rank 33 among top 150 international teams in combustion, electric and driverless disciplines. • 2020 Formula Bharat, Kari Motor Speedway, Tamil Nadu, India - Overall rank 4 out of 60 teams. - Fastest skidpad timings of 4.99 secs. - Finalist of Cost and Business event.Led a subteam of 4 members in research and development of the Perception System of team's first ever Autonomous race car using monocular cameras and NVIDIA AGX XAVIER. - Worked on object detection algoritms using YoloV3 architecture to detect environment landmarks. - Implemented data-augmentation techniques (Scaling and Tiliing) to reduce rigorous data collection. - Tools used: Python, PyTorch, Matlab, Simulink.

Oversaw the entire design, fabrication and intergration of all sub-systems of the car.• Defined and delivered vehicle system level targets by coordinating with fellow department leads.• Performed kinematic and dynamic analysis of suspension system in Matlab.• Implemented structural analysis techniques on system components using Ansys Workbench 16.0 FEA.• Developed a steady state handling model of the vehicle achieving a lateral acceleration of 1.4 g.• Built a robust framework to determine suspension setup parameters by utilising FSAE TTC tire data.• Developed in-house Arduino controlled data logging system using various sensors.• Built an automated gear-shifting system for the power drive (Triumph Daytona 675R) of the race car.

• Implemented an elaborate and accurate approach towards determining the structural stresses of: 1. Chassis members 2. Suspension links 3. Wheel assembly components.• Designed and manufactured an innovative race spec pull-rod suspension system • Designed first of its kind light weight and high performance hub less wheel assembly.• Manufactured complete aerodynamic package from carbon fibre composites which included, - Material sourcing. - Prototyping and testing for bending, shear failure and rigidity. - Mould preparation. - Implementing vacuum bagging technique for optimising weight and performance.