building a network time series foundation model

Analysing working memory and comparing information representation with other forms of memory, during eye tracking.

- Analysing learning speeds of neural networks with similar tasks, to gain insights into learning as a whole.-Used RNNs to model visual perception bias of the brain, and compared it to the neural activity without perception bias‑ to gain insights into evolutionary visual biases. Building a way to visualize the RNN state space and fixed points to be along a ”ring attractor” structure.

A youth group, aiming to end animal abuse and promote animal welfare in our community. We undertake social drives, awareness campaigns and even technical projects to better the animal rights movements in India. Technical projects range from application and web development to product designing - all aimed at reaching an end product, which directly or indirectly improves the quality of life of animals.

- Fixed nearly 10 AI workloads related issues ranging from convergencefailure cases in Habana training/ inference cards, to kernal fixes∙ Was crucial to the debug and validation team while reaching the milestone of Amazon deploying our cards along with EC2 instances∙ Received 2 recognition awards for fixing 2 such crucial bugs, and forsoftware development in the NIC domainArea: AI on hardware, NLP, Computer Vision, and other Workloads debugTech Stack: Python… Show more - Fixed nearly 10 AI workloads related issues ranging from convergencefailure cases in Habana training/ inference cards, to kernal fixes∙ Was crucial to the debug and validation team while reaching the milestone of Amazon deploying our cards along with EC2 instances∙ Received 2 recognition awards for fixing 2 such crucial bugs, and forsoftware development in the NIC domainArea: AI on hardware, NLP, Computer Vision, and other Workloads debugTech Stack: Python, Tensorflow, Pytorch, Keras, C, TPC-C Show less

∙ Led an applied research project for building communication efficient Federated Learning (FL)∙ Architected, and worked on a novel method of using Autoencoders to compress the gradient weight updates from server and collaborator (Edge AI) to reduce communication bandwidth in FL∙ Achieved dynamic and orthogonal compression in the weight updates. ∙ Jointly authored another research paper on a novel divide and conquer approach to FL.Area: Federated Learning | Distributed/ Edge… Show more ∙ Led an applied research project for building communication efficient Federated Learning (FL)∙ Architected, and worked on a novel method of using Autoencoders to compress the gradient weight updates from server and collaborator (Edge AI) to reduce communication bandwidth in FL∙ Achieved dynamic and orthogonal compression in the weight updates. ∙ Jointly authored another research paper on a novel divide and conquer approach to FL.Area: Federated Learning | Distributed/ Edge AITech Stack: Python, Pytorch, Tensorflow, Keras Show less

- Worked on Graphical Neural Networks (GNNs) and it's applications, and modeled a road safety index pertaining to vehicle accidents for the same.- Research Assistant (RA) to the ML and NLP courseworkArea: Data Science, Graph Neural Networks, Machine LearningTech Stack: Python, R, Keras, Pytorch, Tensorflow, Scikitlearn, Matplotlib, Numpy

- Led a research project to achieve Electromechanical System Control using an approach of consecutive Recurrent Neural Networks (RNN).- RNNs were used to model electromechanical systems, and respectivecontrollers- Research paper presented as a part of CCIS proceedings, SpringerArea: Data Science, Recurrent Neural Networks, Control SystemsTech Stack: Python, C++, Tensorflow, Keras

- Built an Android Application handling image processing and text detection and related functionalities by adding to implementations from Google Vision API. Implemented the same onto Android Studios using Google’s ML-Kit. Area: Machine Learning, Android Application DevelopmentTech Stack: Java, ML-Kit, Vision API, Android Studios