Built ELMo model from scratch, capturing contextual language embeddings and enhancing the model’s ability to comprehend nuanced linguistic features, leveraging custom ELMo embeddings to enhance the 4-way news classification task’s accuracy.
Implemented the Contract NLI paper, finetuning BERT & its variants for Natural Language Inference and span identification to reduce the expenses for companies to manually review contracts. Conducted baseline analysis and implemented the multi-task Transformer model and improved upon the paper’s original results by using various techniques such as handling the class imbalance, using the decoder-only transformer model, and unfreezing the top layers of BERT.
Engineered a microservices-based web application designed to enhance a streaming platform by facilitating multiple providers and streamlined subscription management. Conducted a comprehensive analysis of architectural patterns, meticulously comparing the Monolithic and Microservices approach and implemented tailored strategies to optimize the system’s efficiency.
This is a database management system for a fictional drug syndicate, and it stores all the information about the syndicate. The users can write their own queries to get the desired outputs. The program’s user interface is written in Python, and the database is stored in MySQL.
Magnush is a clone of the Bash shell and it is written in C and executed completely using system calls. It supports various features seen in a traditional Bash shell, such as built in commands, directory specific prompt, input/output redirection, piping, and execution of foreground and background processes.
This is a cloud-based application for uploading, listing, and filtering images. It is built upon a microservice architecture with lean services, that was made by building out a CI/CD process that automatically builds and deploys Docker images to a Kubernetes cluster.