Built an automated platform for fine-tuning LLMs with configurable Docker images (1k+ downloads). Fine-tuned Llama 3.1 8b achieving 48% MFU and improving MT-Bench scores by 28%, with load-balanced inference pipeline using vLLM.
Developed a platform for Purdue's Electric Vehicle Club Autonomous Team that takes a combinations of videos and images and uses 3D Gaussian Splatting to create a 3D scene of the area to view and move around in.
Developed a sentiment analysis model leveraging D-Wave's quantum annealer for faster optimization, achieving 85% accuracy on tweet data using an optimized Naive Bayes Algorithm, outperforming classical optimizers by 3%.
Built gpufindr, a platform to compare GPU instance prices across various providers. Developed API and MCP and connected to AI agent to automatically find and rent the best instance for each project from a text input.
For BoilerMake XII: Developed platform to create AI agents on demand and connect them. Won 1st place for: Caterpillar best use of Cloud Computing, MLH Best use of Generative AI, Purdue Innovates: Most likely to become a business
Astrophysics Paper Presentation slides covering the first exoplanet discovery.
Built an ETL pipeline to clean and process 34TB of Reddit text data into 5GB of high-quality training data using CAIOS and CKS filtering systems for machine learning applications.
Authored and published a comprehensive benchmarking toolkit for assessing quantum hardware performance using Grover's algorithm, with standardized metrics for quantum system evaluation.
Part of ARC (Autonomous Robotics Club), designed and built a fully autonomous campus drone delivery system for packages using PX4 flight controller and computer vision.
Simulate rigid bodies with integration for relativity, gravity, and special environments.