• Working under Prof. Lee A. Cooper on Machine Learning for histopathological data.

• Working under Prof. Alan Yuille on Hyperbolic Deep Learning in Computer Vision at the CCVL lab.

• Employed unsupervised learning methods (SCAN, SimCLR, Recall@k surrogate losses, Representation Learning, Style Transform) for vehicle re-identification.• Conducted Quantization Aware Training and Finetuning. The quantized models exhibited competitive performance compared to the original models, with a mAP score of 0.86, a Rank-1 accuracy of 0.91, and a Rank-5 accuracy of 0.97 .• Developed DAFOSNET, a meta-learning based framework recognizing unknown samples from novel classes in… Show more • Employed unsupervised learning methods (SCAN, SimCLR, Recall@k surrogate losses, Representation Learning, Style Transform) for vehicle re-identification.• Conducted Quantization Aware Training and Finetuning. The quantized models exhibited competitive performance compared to the original models, with a mAP score of 0.86, a Rank-1 accuracy of 0.91, and a Rank-5 accuracy of 0.97 .• Developed DAFOSNET, a meta-learning based framework recognizing unknown samples from novel classes in target query sets and handling visual shifts between domains, as well as identifying target outliers under domain shifts. Show less

• Worked under the guidance of Prof. Biplab Banerjee to implement and analyze state-of-the-art Few-shot super-resolution methods such as Meta-SR, MLSR, and DMCN using PyTorch and Tensorflow on remote sensing data.• Involved in development of MAML-SR, a novel meta-learning based GAN for super resolution that outperforms the current state-of-the-art models by 6.5%.

• Worked under the guidance of Prof. Sumeet Kumar on identifying hidden product placements in YouTube Kids videos using Few Shot Learning.• Created modular image pre-processing pipelines and implemented deep learning models such as Matching networks, Prototypical networks, MAML, and S2M2 to create benchmarks for product placement.• Achieved an accuracy of 83.67% using S2M2 on the novel product placement dataset, the highest known result to date.

• Worked on writing ROS middleware for Non-ROS robots.

• Led a team of 15 multidisciplinary engineers to build autonomous multirotor UAVs. • Oversaw all aspects of UAV production from manufacturing to testing, including coordinating with manufacturing teams, reviewing and approving plans and designs, monitoring quality, analyzing test results, and managing budgets and resources. • Applied deep learning based object detection models (YOLOv3, YOLOv5), classification models (VGG16, InceptionV3), segmentation models (UNets), as well as… Show more • Led a team of 15 multidisciplinary engineers to build autonomous multirotor UAVs. • Oversaw all aspects of UAV production from manufacturing to testing, including coordinating with manufacturing teams, reviewing and approving plans and designs, monitoring quality, analyzing test results, and managing budgets and resources. • Applied deep learning based object detection models (YOLOv3, YOLOv5), classification models (VGG16, InceptionV3), segmentation models (UNets), as well as custom architectures, to perform various autonomous tasks on the UAVs.
• Implemented feature-detector algorithms such as FAST, SURF, and ORB to detect stable points on the tracked object 
with respect to the UAV. • During my tenure as the subsystem head, the subsystem secured a world of rank 4 in the IEEE UAV challenge.• All India Rank 1 SAE-AeroTHON.• World rank 2 SAE SAE Aero Design West 2022. Show less

• Worked on various Multirotor UAV and VTOL projects. • Worked with Robot Operating System (ROS) for working with aerial robots.• Worked on making custom Convolutional Neural Networks for object detection tasks.• Worked on quadcopter dynamics and control systems.• Worked on interfacing different F4 and F7 based flight controllers and various microcontrollers like NVIDIA Jetson TX-2, Jetson Nano, Raspberry pi-3, Intel Movidius NCS with sensors and actuators.• 2nd place in… Show more • Worked on various Multirotor UAV and VTOL projects. • Worked with Robot Operating System (ROS) for working with aerial robots.• Worked on making custom Convolutional Neural Networks for object detection tasks.• Worked on quadcopter dynamics and control systems.• Worked on interfacing different F4 and F7 based flight controllers and various microcontrollers like NVIDIA Jetson TX-2, Jetson Nano, Raspberry pi-3, Intel Movidius NCS with sensors and actuators.• 2nd place in India BRICS Future Skills Challenge 2020. Show less

• My work involved mentoring the juniors, giving out software development projects and keeping track of the progress on the projects. • I primarily worked in the domains of Web development, Machine Learning, and Deep learning inside IOSD.

• Webx Studios is on the mission to empower businesses for their IT needs.