Masters of Science - Computer Science

2023 - 2025

State University of New York at Buffalo, NY - USA

CGPA : 3.7/4.0

Bachelor of Technology - Computer Science

2016 - 2020

Kalinga Institute of Industrial Technology, Bhubaneswar - India

CGPA : 3.95/4.0

University at Buffalo - Buffalo [United States of America]

2024 - 2024

Research Aide, Research Foundation for SUNY

• In-Game Voice Chat Moderation: - Enhanced multimodal conversational AI using GPT-4o-audio on Mutox and Detoxy datasets with reasoning prompts in Python, improving safety in reasoning and intent relevance for in-game voice interactions by 25%.
• Moderating Promotional Images for UGCGs using VLLMs: - Automated Python scripts for image retrieval and preprocessing for an in-the-wild dataset consisting of >1K images. Evaluated Vision LLMs with reasoning prompts, achieving 0.95 F1score using MLFlow.

DELL TECHNOLOGIES

2020-2023

Software Engineer - 2

• CarePulse: Architected a distributed ETL data pipeline with PySpark with >100K / day error click stream log records from PostgreSQL, and storing in MongoDB, enhancing data availability by 85%. Developed a stakeholder logging notification system in Java with ELK Stack, reducing errors/week by 1.5x. Architected asynchronous 10+ map-reduce and aggregated services, performed indexing, decreasing the API latency by 80% in Java, and auto-refreshing UX features in React. Integrated LLM, RAG, and RediSearch to extract semantic context in customer feedback, improving CSAT to 78%.
• Dell Online Returns: Revamped the data flow by consuming data from distributed storage queues, Apache Kafka, and RabbitMQ using C#, reducing the processing failures by 70%. Optimized client performance using lazy-loading in React, reducing TTI to 1.83; developed and tested 2 MFEs using JavaScript, increasing scalability and reusability by 2x.
• Dell Order Status: Performed version control, configured using Ansible, maintained CI/CD, deployed using Docker, and increased code quality to 100% in SonarQube using NUnit. Collaborated with stakeholders to build a monitoring alert system using Splunk, increasing CSAT to 56.1%.
• Covaine Portal: Developed scalable COVID-19 vaccine web application with React, Next.js, PKS, FastAPI, PostgreSQL, and Nginx, with 100% booking for 5K. Collaborated on load testing and observability monitoring using OpenTelemetry and JMeter with >1M DAUs.

High Radius

2019

Summer Intern

• B2B AI Prediction BotDeveloped and deployed a B2B AI-based Fintech web app to predict client payment dates of 100+B2B clients using JavaScript, Java, Python, Flask, and AWS. Collaborated with Senior Data Scientists to research, analyze, and gather requirements to clean, preprocess, and conduct feature engineering on the past dataset of B2B customers. Implemented a supervised regression model with 0.2 RMSE and 0.4 MSE for the prediction of the payment date for B2B clients.

ClimateSyncAI

2023 – 2022

Tech Stack

• Clustered cities into 13 segments using K-Means based on average temps; trained per-cluster SARIMAX models for improved fit.
• Built Flask UI with Bootstrap to visualize forecasts, trend lines, and anomaly flags; Dockerized for easy deploy.
• Achieved lower MAE vs baseline ARIMA; added CI hooks for quick retraining on new data.

Source Code

NYRailSyncDB — Geo-Distributed Railway Database

2024

CockroachDB (Postgres syntax), REGIONAL BY ROW, 3-node Docker Compose

• Modeled trains, routes, bookings, and maintenance with normalized ER schema and UUID keys.
• Used REGIONAL BY ROW for data locality, improving read latency and regional compliance.
• Deployed HA 3-node cluster via Docker Compose; verified ACID transactions under node failure.

In-Game Voice Chat Moderation — Multimodal Safety

2024 – 2025

WavLM & DistilBERT classifiers, LLM evaluation harness, FastAPI test bench

• Built FastAPI framework to evaluate audio toxicity classifiers with precision/recall/F1 and confusion matrices.
• Ran Fleiss’ Kappa for annotator agreement on multi-label classes; automated annotation bootstrapping.
• Integrated prompt-based LLM evaluation; tracked experiments with MLflow for reproducibility.

TradeFlow — Multi-Strategy Trading Bot

2025

FastAPI, RabbitMQ, Redis Pub/Sub, MongoDB, Streamlit dashboards

• Orchestrated 12+ concurrent strategies using RabbitMQ for order pipelines and Redis for LTP streams.
• Persisted trades & tick data to MongoDB; built Streamlit dashboards for PnL and drawdown analytics.
• Containerized services and defined health checks for resilient multi-container deployments.

Sketch-to-Wireframe — Chrome Extension

2023

CNN Autoencoder generating HTML wireframes (divs, inputs, checkboxes) on AWS

• Processed hand-drawn UI sketches to produce semantic HTML skeletons with high element detection accuracy.
• Packaged as a Chrome extension; supported export and quick edit workflows.