Here's a small introduction about myself...
Recently graduated from Stony Brook University, New York completing my Master's in Computer Science. Techie, a Polyglot , I love exploring and fiddling with new technologies in an attempt to solve the hot computer science problems. I live by the words of Steve Jobs, “Stay hungry. Stay foolish.” and approach everything I do with this in mind to guide my curiosity and goals
With about 5+ years of full time experience, working on enterprise B2B apps in the domains of Payments and Loyalty services, I'm a seasoned professional with experience in Java, Python, Javascript and DevOps/Cloud computing. Interested in working on applications at the intersection of Data Analytics and Machine Learning.
Expertise: Data Structures and Algorithms, Supervised Learning, Problem solving and Web programming.
My work during Fellowship at Hasura.
Implementation of Social OAuth
Greyatom Capstone Project
Performed EDA on dataset of 1.7 million records for an Hardware store.
Prediction using Regression Techniques applied on Ames Housing Dataset on Kaggle.
Implemented Skip-Gram model and utilised NCE Loss function to train word-embeddings.
Using Left-Arc, Right-Arc and Shift operations. Designed own Dependency Parser, a custom Loss function and performed experiments by tuning hyperparameters.
Implemented BiDirectional GRUs with Attention for Relation Extraction task. Improved over Basic model by designing an Ensemble network of CNNs with LSTM.
Developed on existing approaches to implement a Unified model using Bidirectional LSTM, evaluated on ROUGE scores.
Scene Stitching using SIFT features with Laplacian Pyramids and object recognition using HOG.S
CNN Models (Scratch v/s Pre-trained AlexNet and VGG16) for Scene Recognition in PyTorch.
Estimating the 3D pose of a person given their 2D pose.
Assessing impact of Novel Coronavirus on Uber/Lyft business in New Jersey,USA using Probabilistic models and Time series forecasting.
D3 Dashboard on Novel Coronavirus to assess its impact on the world
A new factor for passwordless authentication using ultrasound for data transfer
The end goal of this research and implementation was to make charts more sophisticated, interactive and accessible to visually impaired users via text-to-speech..