Parth Kotwal

About Me...

Hello, I'm Parth, a Computer Science student at the University of Washington, Seattle, with a fiery passion for data science, machine learning, and using CS to positively impact communities. I am also super interested in exploring how ML can be applied in the context of astrophysics (black holes, particle physics, time-domain) and security (malware detection, adversarial ML), or just issues in my everyday life.

In my spare time, I enjoy:

Playing Red Dead Redemption 2 or FIFA

Following my favorite football team FC Barcelona

Reviewing restaurants, cafes, and bakeries on Beli (please follow me)

Listening to any music, from Travis Scott to Vishal Shekhar

Feel free to explore my projects and get in touch if you'd like to collaborate or learn more about my work!

Experience

Machine Learning Intern

miniOrange • July 2024 - September 2024

• Constructed and trained a decision tree-based malware detection system (94% accuracy, 92% weighted precision) using feature engineering on a dataset of 20K+ samples; benchmarked and tuned via grid search
• Deployed a k-means clustering pipeline on 10M+ login records for anomaly detection in ISP authentication systems; contributed to integration planning for adaptive security infrastructure
• Documented model training and deployment steps in to support future iterations by the artificial intelligence team

Financial Data Science Intern

Stonehaven Capital Management • September 2023 - November 2023

• Supported the development of a data pipeline to assess IPO success, using competitor performance metrics and historical market data
• Automated preprocessing of 6+ financial datasets via Yahoo Finance APIs, creating visualizations for each dataset

Projects

QueQueue

Django Vue.js Scikit-learn AWS (Elastic Beanstalk, S3, RDS) PostgreSQL

Building a cross-platform Spotify and Apple Music queue manager with Django and Vue.js, allowing users to save, export, and restore sessions across devices. Created a ‘suggest’ feature using audio feature clustering and cosine similarity for mood-based recommendations. Deploying the backend using AWS Elastic Beanstalk, S3 for media storage and RDS as the primary relational database

Transient Detection with braai CNN

PyTorch CuPy NumPy Scikit-learn

Engineered a data pipeline to classify real/bogus astrophysical transients in Zwicky Transient Facility image triplets. Reimplemented each of the original CNN’s layers from scratch using CuPy and NumPy, optimizing convolutions and pooling on GPUs with im2col for 99.97% reduction in training time. Benchmarked CuPy CNN against a PyTorch counterpart and classical models, achieving ROC AUC=0.91 and F1=0.82

Model United Nations Northwest SMS Admin

Flask JS PostgreSQL Tailwind CSS

Established a Flask-based SMS automation system using Twilio and PostgreSQL, allowing real-time announcements for 2,100+ conference participants across four major student conferences. Developed a secure admin dashboard with authentication, bulk contact management, and data validation for seamless user operations. Replaced fully manual staff-dependent workflow, crucially accelerating announcements via multithreading techniques

Automated Stellar Classification

Scikit-learn Pandas NumPy Matplotlib

Trained and evaluated five machine learning models on stellar data (luminosity, temperature, absolute magnitude) to classify stars into 7 spectral types (Morgan–Keenan system). Optimized a random forest model to 95% accuracy and 96% precision; visualized key features to improve interpretability and connect machine learning with scientific modeling

Skills

Languages

Python JavaScript R Java C++

Frameworks & Libraries

PyTorch Scikit-learn Django Flask Vue.js Pandas NumPy Matplotlib Tailwind CSS

Tools

Git Jupyter AWS (Elastic Beanstalk, S3, RDS) GCP (Translation API) Twilio PostgreSQL Visual Studio Code Eclipse