Portfolio

Journey so far...

2024

University at Buffalo

Research Assistant

Responsibities -

Compile dataset to classify dialogue utterances into 6 emotion labels
Fine tune roBERTa base and large LLMs for Augmentative and Alternative Communication NLP module
Generate emotion aware responses to assist speech impaired users

2023

University at Buffalo

- The State University of New York

Master of Science

Computer Science and Engineering

GPA: 3.38/4.0

Coursework - Analysis of Algorithms, Data Models and Query language, Information Retrieval, Data Intensive Computing, Modern Networking Concepts, Machine Learning, Computer vision and Image Processing, UI/UX, Project Development, Project Management.

2022

HSBC Software Development

Software Engineer

Responsibities -

Worked on critical tier-0 business application
Built full-stack application in Java, Angular, Spring Boot
Utilized Elasticsearch for JSON datastore and logging solution
Assisted in re-architecturing from monolithic to microservice
Optimized CI/CD deployment operations

.

2020

Pune Institute of Computer Technology

Bachelor of Engineering

Information Technology

GPA: 9.21/10

Coursework - Data Structures and Algorithms, Operating Systems, Computer Networking, Machine Learning ALgorithms, Database Management Systems, Data Science and Big Data Analysis.

My work examples

Novel Chat Bot

Developed and executed with microservice architecture a conversational Q&A chat-bot implemented with Retrieval Augmented Generate (RAG) to answer questions from 49,000 documents indexed in Solr Used Apache Solr, Python, Google Cloud Run, React.js, Tailwind technology stack.
Implemented microservice architecture to pre-process query, classify text label, fetch relevant documents, and generate appropriate answer. Utilized DialogGTP model and Google Vertex AI.

Caption CraftLink

Created a application that generates context aware captions for images to post on social media. Constructed UI with React.js and backend in Python, Flask, Gunicorn.
Built on top of the Blip Image Captioning model and Google Vertex AI (PaLM 2). Containerized and hosted on Google Cloud Run.

Data Volume Reduction (with IBM)Link

Designed, documented and executed a containerized application that enables users to filter Geo-Spatial and Exif Image dataset based on metadata. Developed modules with Django-Rest framework with Next.js and Python.
Conceptualized a standard API and filter query to expose metadata and selectively download only required subset of data.
Built UI, with React js, to select filtering criteria and download reduced dataset in a .zip/.tar.gz format. Productionized with IBM Cloud Engine.

Breast Cancer AnalysisLink

Charted and built a portal using Flask and Python, to predict mortality of breast cancer patients from relevant factors. Trained 5 machine learning models that achieved an accuracy of 85% to predict outcomes.
Investigated datasets, performed Data Cleaning and Exploratory Data Analysis and visualized results to identify trends and interpret factors that directly or indirectly impact the tumor size and survival span in breast cancer patients.

Server Chat Application

Orchestrated client and server components of a text chat application,consisting of one chat server and multiple chat clients over TCP connections. The server facilitates communication between clients.
Programmed TCP connections using standard C libraries and socket programming. Performed “select” subsystem for I/O multiplexing and coordinating multiple sockets.
Modeled features like login, send, broadcast, block, unblock for clients and statistics, relay events for server.

Audio Speech Emotion Recognition

Generated a lightweight interpretable machine learning model in python to recognize emotion of audio speech.
Examined and compared the effect of hand-crafted audio signal features, extracted using Librosa library, on multiple classifiers to create an ensemble model for emotion recognition.

Student-Mentor Bridge Portal

Created an engaging communication platform for mentors and students with features, including leave applications, meeting scheduling and performance tracking; along with admin side features like allotment of students to mentors. Implemented advanced features like relocating/updating students assigned to mentors and tracking their history.
Designed the database schema in 3NF form. Implemented complex SQL procedures and triggers for optimizing query performance.
Used Java, JDBC, Servlets, MySQL for backend and HTML, CSS, Bootstrap for frontend.

I Researched

Impact of Lightweight Machine Learning Models for Speech Emotion Recognition

Springer - August 2021

The comparative research of single and ensembled Machine Learning models for recognizing the kind of motion from a verbal utterance was explored. The study demonstrates how Machine Learning models with lower resource usage potentially outperform Deep Learning models.

Survey on Information Retreival Accuracy in Biomedical Systems

Elsevier SSRN series - May 2020

Surveying Architecture of Medical Information Retrieval (IR) System, IR Systems/Models, Traditional IR models and summary of IR along with its methodologies and its importance in Biomedical IR Systems is done in the paper.