Machine Learning Research Assistant at University of South Florida (2026-01 – Present)
- Annotated 20 high resolution sattelite images with Roboflow to ensure clear input data for 3000 patches.
- Engineer a computer vision algorithm with CNN, TensorFlow and Keras to identify and count ships.
- Created a data pipeline with Python to reduce processing overhead and improve analytical efficiency with 5 day a month data from 2021 to 2023.
- Documented code and version changes in GitHub to maintain traceability and support team collaboration.
Medical Imaging AI Classifier (MRI) at University of South Florida (2025-12 – 2025-12)
- Created an early detection system using Image Processing, Keras, TensorFlow to prescreen MRI images for cancer possibilities.
- Developed a Convolutional Neural Network (CNN) to detect and classify cancer types from MRI scans, achieving a validation accuracy of 82%.
- Implemented image preprocessing techniques using pandas (normalization, resizing) to optimize dataset quality for model ingestion.
Data Engineer at University of South Florida (2025-09 – 2025-09)
USF Health Hacks Winner (BCI Track)
- Processed brain EEG signals in Python and Whisper API 8 times per second and processed signals into commands using C++ and Arduino IDE to control motor hand.
- Created a wireless version with Python, signaling through Bluetooth for further distance connection.
- Created a cheaper version of brain-controlled prosthetic 30% cheaper than the market prototypes ($750).
Systems Analyst at University of South Florida (2025-08 – 2025-11)
Shop Now e-commerce platform
- Designed and implemented an Oracle 23ai OLTP database for a simulated e-commerce platform with 10 sample records, 20+ attributes, and fully defined relational schema documentation.
- Created 6+ relational tables, ERDs, data dictionaries, and DDL/DML scripts to standardize data structures and support end-to-end database setup.
- Wrote 15+ SQL queries using joins, subqueries, aggregations, DDL, and DML to validate relationships, load sample data, and test business reporting logic.
- Improved query efficiency by building data cubes that reduced lookup time to 0.005 seconds.
