Fine-tuned a transformer-based LLM on MPT-30B from MosaicML to specialize in scientific computing, physical engineering, and automated data analysis. COSMO ingests datasets from AWS S3 and uses Apache Spark for scalable training. It generates insights, summaries with raw data, time series, and sensor feeds. Built on Databricks for data prep and deployed with Kubernetes, COSMO supports domains like aerospace telemetry, materials modeling, and satellite operations.

Created 'Pika Run' an simple open-world mini-game using featuring a text-based engine that simulates dynamic world exploration through procedural room generation, interactive gameplay, and strategic navigation challenges.

Developed a full-stack web application with a 3D UI for exploring orbits around the globe & with AI enhanced orbital prediction using LSTM tracking that learns in real time. Integrated a PostgresSQL backend by AWS for satellite data and a React frontend for dynamic visualization/interaction.

Developed a predictive model to analyze and infer properties from over 3,000 stellar spectra of red giants in the APOGEE survey using Python.

Created a version control system similar to Git, allowing you to track changes in your codebase over time

Developed a sophisticated machine learning model using PyTorch to classify galaxies based on shape, utilizing images from the Sloan Digital Sky Survey and Galaxy Zoo project.