- Developed a Physics- Informed LSTM (PILSTM) model for forecasting Snow – Water Equivalence (SWE) and reservoir inflow using PyTorch, achieving a Nash – Sutcliffe Efficiency (NSE) of 0.837 for SWE and 0.82 for reservoir inflow.- Designed a comprehensive data pipeline to transform 1000+ hydrological and meteorological data points into LSTM inputs.- Utilized Savio HPC for comprehensive hyperparameter optimization of PILSTM models, reducing MSE by 25%.

- Built LSTM neural networks using PyTorch to predict key core performance metrics and distributions of pebble bed nuclear reactors. Achieved MSE of .016 at stable power and .171 at varying levels of reactor power. - Generated training data via Monte Carlo simulations and built data pipeline to convert 40k features into LSTM model inputs.- Optimized LSTM model by identifying 10 key isotopes from ~500 using high correlation/low variance feature removal.

- Managed 3 bootcamps with 10+ instructors and 200+ students, optimized outreach program to students and saw a 30% increase in student attendance- Ran community events and saw organized socials, saw a 45% increase in satisfaction from students and instructors

- Led 2 bootcamps with more than 100+ students combined, optimized lesson plan and curriculm and saw a 50% in satisfaction from students

- Taught a nonprofit data science to bootcamp for 50+ students, teaching them basics of data analysis, data visualization, and data science (Data 8 and Data 100 curriculum)- Created curriculum used by over 200+ students over the past years, saw a boost in student satisfaction by 25%

- Analyzed 4M+ tweets using ALBERT for sentiment analysis on global social media, identifying key patterns in COVID-19 misinformation to enhance public health understanding. - Developed data extraction scripts and employed NLP techniques across 30+ languages, improving data accuracy and efficiency in analyzing vaccine acceptance and COVID reporting. - Enhanced NLP models using Python and fine-tuned a BERT-derivative LLM model to identify misinformation trends, utilizing techniques like sentiment analysis and topic modeling.

- Assisted students in a variety of topics consisting of probability, statistics, coding, and classification models.- Taught over 50 students for over 20 hours in a 5 month span.

- Led a team of 5 in analyzing COVID-19 trends through global vaccination data, engineering a dashboard that would provide real-time COVID-19 statistics delivery within 30 seconds to enhance data-driven decision making.- Implemented a hybrid ARIMA + SEIR predictive model with Scikit-Learn and PyTorch, attaining 67% accuracy in forecasting next-day COVID-19 numbers

- Worked with a group of 5 to develop a web app to simulate field and robot actions for the PiE Robotics Competition to help competitors test code without field.- Utilized JavaScript, Bootstrap, Web Workers, GitHub, and HTML Canvas to create a test field with different field configurations as well as implemented line following sensors, interactable objects, and an internal IDE

- Led a group of 3 students to build a competition ready beetleweight battlebot. Utilized CNC Machines, Waterjets, and Solidworks CAD Modeling Software to construct the robot.- Built a completely autonomous sumobot. Utilized sensors to implement line following, spatial awareness, and basic decision making. Used Fusion 360 to model the robot and programmed it using C.