Hi, I am Nirbhay Sharma. I am currently pursuing M.Tech in Artificial Intelligence (AI) from Indian Institute of Science (IISc) Bangalore. I have recently completed my B.Tech in Computer Science and Engineering (CSE) from IIT Jodhpur. I am a Machine Learning enthusiast and hold a good experience in Pytorch and Python I take a keen interest in reading new state-of-the-art research papers and implementing them fascinates me. Apart from that, my hobbies include exercise and yoga so I can keep myself fit to learn the skills :)
Education
M.Tech Artificial Intelligence, Indian Institute of Science (IISc) Bangalore, CGPA: 8.7 /10, AIR-6 GATE DA
B.Tech Computer Science and Engineering, Indian Institute of Technology (IIT) Jodhpur, CGPA: 8.97 /10
Class 12, PCM, Dehradun Public School, Percentage: 96.4
Class 10, SD Public School, CGPA: 10 /10
Publications
An Extremely Lightweight CNN Model For the Diagnosis of Chest Radiographs in Resource-constrained Environments, International Journal of Medical Physics 2023 , [Paper]
Aggregation-Assisted Proxyless Distillation: A Novel Approach for Handling System Heterogeneity in Federated Learning, International Joint Conference on Neural Networks (IJCNN) 2024 , [Paper]
Proposed a novel FL Framework FedAgPD to simultaneously handle model and data heterogeneity
Leveraged Deep Mutual Learning at Client and Aggregation followed by Gaussian Noise based data free
distillation at the Server, eliminating need of proxy dataset or GAN's
FedAgPD achieved 2x better performance compared to SOTA FL algorithms like FedDF, FedMD, Kt-pfl
IIT Jodhpur
Extremely Lightweight CNN for Chest X-Ray Diagnosis
June2021 - May2022
Supervisor: Dr. Angshuman Paul
Designed a novel Lightweight CNN model (ExLNet) for the abnormal detection of Chest Radiographs
Fused Squeeze and Excitation blocks with Depth-wise convolution to create DCISE layer as a component of
ExLNet, which outperforms SOTA models like Mobilenet, Shufflenet on NIH, VinBig medical datasets
IIT Jodhpur
Cell Detection and Classification
August2022 - May2023
Supervisor: Dr. Angshuman Paul
Detected and classified cells data sample into necrotic and apoptotic cells
Finetuned various SOTA object detectors such as YOLO, SSD, RetinaNet, DeTR
Achieved remarkable results using DeTR with a Mean Average Precision (MAP) of 40.0
Worked on Split Neural Network ML paradigm and Splitted Mask-RCNN, FCN_Resnet50, YOLOv5 models for
Instance segmentation, segmentation, face detection tasks
Implemented Autoencoder model for efficient image compression to latent space and setup Pysyft to
communicate latents from Jetson Nano to GPU server, preserving data privacy at Jetson Nano
Faaya Astu
Content Generation, Avatar Generation
July2023 - July2024
Supervisor: Sahil Bajaj
Trained Stable Diffusion ControlNet models on Lineart and Colorbox control on VastAI GPU instance and
deployed them on RunPod for more flexibility and control on print generation
Trained Low Rank Adaptation (LoRA) with Kohya_SS for custom face and background generation
Experimented with custom ComfyUI workflows with integrated ControlNet, LoRA, InstantID models
Containerised ComfyUI with Docker and deployed them as Serverless Endpoints on RunPod and exposed endpoint
APIs to AWS Lambda to create APIs for APP using AWS API gateway
Projects
Semester-1 @ IISc | ML/DL
Tech: Numpy, Python
The repo contains the first semester assignments and PYQs for STOMA, LAA, DSA, CMO
Image Captioning using Detection Transformer (DeTR) | ML/DL
Tech: Pytorch, Transformers, Python
Implemented modified DeTR from scratch in pytorch for image captioning task. Trained DeTR on Flickr30k dataset for 500 epochs and achieved a BLEU score of 57.36 on Flickr8k dataset
Vision Transformers Implementation | ML/DL
Tech: Pytorch, Vision Transformers, Python
Implemented 11 SOTA research papers on vision transformers variants like Swin Transformer, Pyramid ViT, Convolution ViT etc. for Image Classification from scratch in pytorch
Regularizing Federated Learning via Adversarial Model Perturbations | ML/DL
Tech: FL, Pytorch, Python, AMP
The project aims at implementing the state of the art methods for Federated Learning (FL) like scaffold, FedNTD, FedProx, FedAvg and regularize the client using adversarial model perturbations to reach flat minima.
CNNAlgos-Comparison | ML/DL
Tech: Python, Pytorch
The project consists of various deep CNN architectures (coded from scratch) on Retinal eye disease dataset (kaggle), and performed a comparative study among these deep architectures
Image Colorization | ML/DL
Tech: Python, Pytorch
The project aims at implementing Pix2Pix GAN architecture from scratch on RGB and LAB image format, to convert a black and white image to colored image
PRA-Visulaizer | web-dev
Tech: React, Javascript, HTML/CSS, JSX
The project consists of visualization of various page replacement algorithms such as (FIFO, LRU, OPR) etc. given number of frames and demand pages, the app can visualize how various algorithms handle page replacement.
Email-app | web-dev
Tech: Django, Python, Sqlite3
the project is build using Django framework in python and the database used is sqlite3, the project implements the functionalities of email application (inbox, compose, sent, deleted mails) etc.
Mask-NoMask detection | ML/DL
Tech: Python, Pytorch, Flask, OpenCV
the project is build using pytorch library and the final trained model is then used for real time detection using openCV and also the testing can be done from web application build using flask
