About Me
Hi, I'm Prasad. I'm pursuing my Master's in Data Science, where I focus on machine learning and building intelligent systems. I enjoy designing and developing websites, Android apps, and building APIs, and I'm interested in creating clean, minimalistic solutions that solve real problems.
Outside of tech, I'm passionate about photography, especially astrophotography and landscapes. I enjoy planning night shoots, working with long exposures, and refining images in Lightroom. I also like biking, playing guitar, and reading nonfiction, which keeps me curious and always learning.
Summary
On the application side, I build backend systems using FastAPI and Flask, and create user interfaces with React and Tailwind. I work with databases such as PostgreSQL and MongoDB to design reliable data systems. I deploy projects using AWS, Docker, and CI/CD pipelines to ensure smooth, production-ready performance.
I've worked on ML projects using PyTorch and BERT for NLP tasks, including spam detection and topic modeling with BERTopic. My experience also covers distributed training and hyperparameter optimization with Optuna.
I've also explored search and retrieval systems using FAISS and Elasticsearch, and recently started experimenting with LangChain for LLM-based applications, along with blockchain-based projects to better understand decentralized system design.
Experience
Data Analyst Intern (NLP)
Oct '24 - PresentWisconsin School of Business
Engineered a batch ETL pipeline in Python and SQL to ingest and normalize structured hotel reviews, enabling downstream NLP analysis at scale.
Stabilized model performance across demographic segments by implementing a robust SpaCy text preprocessing pipeline, reducing variance in F1 scores.
Designed a modular topic modeling service using Sequential LDA to automate the categorization of customer sentiment in reviews.
Improved topic coherence by 7% by transitioning from Sequential LDA to BERTopic with transformer-based embeddings.
Android Development Intern
Jun '21 - Aug '21StoccGuru
Integrated the Android XML front-end with a Python Flask backend via REST APIs, achieving response times under 250ms.
Developed a comprehensive automated testing suite using Postman for unit and integration tests, significantly reducing API failure rates in production.
Delivered core backend modules across 5 sprints using Agile/Jira methodologies, ensuring 100% on-time release.
Education
University of Wisconsin - Madison
M.S. Data Science
2024 - 2026
University of Mumbai
B.Tech. AI & DS
2020 - 2024
Publications
LightGBM and Gradient Boosting for Optimizing Shipment Mode in Pharmaceutical Supply Chains
Springer Singapore
Read PaperProjects
Tydos
Config files for my GitHub profile.
Credit Risk Scorer
End-to-end ML pipeline for predicting loan payback, featuring data prep, model training, MLflow tracking, FastAPI deployment, monitoring, and orchestration.
Pharmaceutical Supply Chain Optimisation
A machine learning pipeline for predicting pharmaceutical shipment delivery risk. The system processes supply chain data, engineers features, and trains models to identify shipments at risk of late delivery — enabling proactive supply chain management.
GPT 2
GPT-2 built from scratch in PyTorch. Trained on WikiText-103 in ~25 mins/epoch on an A100.
Portfolio
Full-stack personal portfolio built with Next.js 14 (TypeScript) and FastAPI (Python) - features a live photo gallery backed by Supabase, GitHub project integration, and a static resume section. Deployed on Vercel.
Antares
Built an AI system for intelligent PDF retrieval and question answering.
Photography
Let's build
something.
Let's connect over e-mail!