* Delivered 10+ in-house Machine Learning lectures - Neural Networks, Autoencoders, Transformers, CNNs, GANs, and Optimization Techniques, to diverse student audiences* Led a 10-person team in designing and developing a React/Bootstrap-based website, implementing Agile workflow practices for optimal collaboration and project execution* Managed six student-led projects, including the development of a Reinforcement Learning chess player and a sentiment-based stock predictor.* Organized large-scale university events such as the GenAI Town Hall (150+ attendees), fostering partnerships with prestigious labs including Harvard and UVA LinkLab.

• Trained and deployed Probability of Default Model in SAS, Python, and SQL on quarterly loan data for client• Optimized existing data pipeline to use pandas, NumPy, and Multiprocessing – achieved 380% reduction in time• Built novel Post-Hoc Model Explanations algorithm to estimate feature contributions within financial model• Estimated Shapley Values by clustering highly correlated features with K-Means to then compute Owen Values

• Trained transformer (BART) model with HuggingFace and Pytorch to summarize PDF’s• Built Flask web application and summarization pipeline to upload, parse, and summarize PDF’s and extract graphics• Prototyped on-device, ChaptGPT-like Q/A platform allowing users to upload and question their documents without OpenAI• Fine-tuned open-source Large Language Models and built vector database with ChromaDB

• Developed feature-matching algorithm to detect overlap video frames and stitch images with OpenCV (Python/C++)• Reduced run-time by 200% via multi-threading, mutexes, and coding safe queues/maps to prevent deadlocking• Designed an end-to-end, robust system architecture transferring data between Python and C++ modules • Constructed communication pipeline between web application and hardware via Flask and Sockets

• Initialized and administered NetSuite ERP for a custom apparel startup and collaborated sales to decrease workload• Automated tasks by building custom plugins, web pages, forms, etc. with JS/HTML/CSS and the SuiteScript API

• Designed end-to-end American Sign Language recognition models for the Deaf and Hard of Hearing• Trained SVMs, decision trees, and MLPs on OpenPose (pose estimation) data of sign language• Trained RNN model on IMU data from smartwatches to translate gestures to speech

• Created an online store and platform for students K-12 to share media and learn about online safety• Designing a separate UI for children and adults (VueJS, HTML, CSS)• Designed a user system allowing users to store media submissions (Django and SQL)