CV

PhD

University of California, Riverside, USA

MS

University of Washington, Seattle, USA

Research Assistant at CTRL

University of Washington, Seattle

Research focused on multimodal data fusion and human detection for search and rescue missions

• Achieved a remarkable 40% improvement in fusion quality of multimodal data compared to existing methods by designing and implementing a comprehensive pipeline based on Generative Adversarial Networks (GANs) with cross attention to fuse misaligned thermal and visual images.
• Developed a deep learning model using an autoencoder, with mutual cross attention, a mechanism that allows the model to focus on the most relevant features of different modalities, to fuse data from thermal and visual images for human detection in search and rescue missions.
• Accurately annotated human detection bounding boxes on the Wilderness Search and Rescue Dataset (WiSARD) using Supervisely, driving the development of robust computer vision models for precise human detection and tracking to help in search and rescue missions.

Senior Research Engineer

Mahindra Research Valley, India

Research and development in automotive engine optimization and emission control systems

• Devised a neural network based DOC (Diesel Oxidation Catalyst) plant model using PyTorch & researched on DOC thermal mapping to control and optimize fuel required to regenerate the DPF to increase the fuel efficiency for the engine.
• Optimized the combustion parameters for Diesel Engines by working on Design of Experiments (DOE) to maximize the temperature of exhaust gas entering DOC (Diesel Oxidation Catalyst) to generate a suitable environment for DPF (Diesel Particulate Filter) regeneration.
• Improved regeneration interval (+100Kms) and fuel efficiency by carrying out winter validation of BS6 SUV at Manali, India. Effectively analysed the collected data using Python and other tools and gave further inputs to solidify the calibration.

3D‑AI‑Camera‑Aided VTOL‑Multirotor Vehicle Control & Landing (Capstone Project With INSITU ‑ A Boeing Company)

AI‑camera enabled controller with depth perception and object detection for multi‑rotor VTOL vehicle landing

• Built an AI‑camera(OAK‑D) enabled controller with depth perception and object detection to land a multi‑rotor VTOL (Vertical Take‑off and Landing) vehicle into a moving confined space.
• Worked on an object detection algorithm based on YOLO, to enable the multi‑rotor VTOL vehicle to land into the moving platform of 1 square meter area, within 60 sec of being detected.

ASL Detection and Letter Prediction

Real-time American Sign Language recognition using computer vision

• Developed a program to recognize hand gestures in 3D space using a single low‑resolution camera to display American Sign Language (ASL) as English Letters in real‑time.
• Trained the model on YOLO and used the trained model with OpenCV, to recognize the hand gesture and predicting ASL letter.

Finalist in Hollomon Health Innovation Challenge 2024

Finalist in Alaska Airlines Environmental Innovation Challenge 2024