Projects

NYC Taxi Demand Prediction
Forecasted taxi ride demand using weather, event, and flight data.
- Dockerized pipeline & deployed with MLflow on Chameleon Cloud
- Improved prediction accuracy by 28%
- Reduced retraining time by 40% using Airflow orchestration
Python
XGBoost
Airflow
Docker
MLflow
View Code

NYC Crime Risk and Route Prediction
Machine‑learning powered crime‑risk forecasting with real‑time routing.
- 95% accuracy across 300+ neighborhoods
- Reduced route risk exposure by 25%
- Processed 500 000+ crime records
- 40% reduction in API response time
Python
Streamlit
Flask
Scikit‑learn
JavaScript
View Code

Optical Character Recognition App
OCR system using OpenCV and Tesseract for high‑accuracy text extraction.
- 95%+ recognition accuracy on 10 000+ images
- 30% boost in OCR accuracy
- 80% reduction in manual entry
- 25% reduction in errors
Python
OpenCV
Tesseract
Flask
HTML/CSS
View Code

Faculty Feedback System
Automated sentiment‑based review system with real‑time visualization.
- 500+ evaluations automated per semester
- 50% faster faculty review cycle
- 30% improvement in UI experience
Python
Power BI
Node.js
SQL
HTML/CSS
View Code

SYMBI‑MART E‑commerce Website
Unified on‑campus stores with AI recommendations and improved UX.
- Boosted order volume by 50%
- Integrated RoBERTa for smarter search and listings
- Optimized inventory management for 3 stores
Node.js
NCF
RoBERTa
JavaScript
HTML/CSS
View Code

Netflix-Inspired Streaming Platform
Recreated the Netflix experience with responsive design, seamless navigation, and dynamic media playback.
- Implemented React.js and Firebase for a full-stack SPA
- Enabled user authentication with real-time content updates
- Integrated TMDB API for dynamic movie recommendations
Node.js
NCF
JavaScript
HTML/CSS
View Code

Document Classification & Extraction
Automated NLP pipeline for document sorting and key‑info extraction.
- 92% classification accuracy on 50 000+ documents
- 40% improvement in classification precision
- 60% faster document retrieval
- 75% drop in manual overhead
Python
TensorFlow
BERT
View Code