Projects

BIONIC is a web-based platform that delivers AI-driven rehabilitation for prosthetic users. A Streamlit/HTML/CSS/JavaScript front end streams video to OpenCV and MediaPipe, which isolate prosthetic limbs and capture movement data. Custom-tuned Google Gemini models analyze that data to generate personalized exercise recommendations, while an NLX Dialog Studio chatbot provides scheduling, guidance, and support via text.

• •Integrated real-time video processing for movement analysis
• •Built chatbot for user support and guidance
• •Implemented data analysis pipelines with Google Gemini

Appoint AI is a web-based triage system that assigns medical urgency scores from patient-submitted symptom descriptions. A React/HTML/CSS/JavaScript front end captures input and displays results. Behind the scenes, Python powers a neural-network model (trained on synthetic symptom data) to score urgency, while a scheduling engine auto-books earliest appointments for highest-risk cases.

• •Developed neural network model for urgency scoring
• •Created automated scheduling engine
• •Built Streamlit front-end interface for patient symptom input and results display

Began developing a Learning Management System (LMS) platform for TechBridge that would allow them to upload their own curriculum and allow other orgs to upload their curriculums as well. Process included understanding the code given and how to update and change important functions of the site.

• •Analyzed an existing LMS to understand curriculum management workflows
• •Evaluated how organizations could upload and manage course content
• •Reviewed user and organization management logic within the platform

A sentiment-analysis pipeline was adapted to classify 115,661 tweets as racist, sexist, or nonoffensive. Text is preprocessed in Python (tokenization, stop-word removal, normalization) and vectorized with pretrained embeddings (BERT). A fine-tuned neural model (TensorFlow/Keras or PyTorch) then assigns labels, with performance validated.

• •Fine-tuned BERT model for bias classification
• •Built preprocessing pipeline for tweet analysis
• •Achieved high accuracy in sentiment and bias detection