Fashion Recommendation System
Problem: Users had no personalized way to discover fashion items suited to their taste.
Approach: ML-based recommendation engine, deployed as a Streamlit web app.
Result: Boosted user engagement and sales.
Research Paper Title Generator
Problem: Researchers had no automated way to turn an abstract into an academic-style title.
Approach: Fine-tuned a BART Transformer model for abstractive title generation from abstracts.
Result: Generates academic-style paper titles directly from abstract text.
Loan Approval Automation
Problem: Manual loan-eligibility review was slow and inconsistent.
Approach: ML eligibility-prediction model backed by MongoDB/PyMongo, deployed as a Streamlit web app.
Result: Automated loan-eligibility predictions end-to-end.
Uber Ride Data Analysis
Problem: Raw ride data wasn't structured for quick business decision-making.
Approach: Analyzed Uber ride data and built an interactive dashboard with charts and visualizations.
Result: Enabled clearer, data-driven business decisions.
Diabetes Heart Monitor System
Problem: Needed an accessible way to combine diabetes-risk prediction with real-time vitals monitoring.
Approach: Random Forest Classifier for diabetes prediction plus real-time heart-rate monitoring via Pulse Sensor + Arduino; built both the AI and the web deployment.
Result: 83% diabetes-prediction accuracy.
Microsoft Stock Price Prediction with Machine Learning
Problem: Needed a data-driven way to forecast MSFT price movement instead of manual technical analysis.
Approach: ML model applying technical analysis over 2013–2021 historical price data.
Result: Produces forward price predictions for Microsoft (MSFT) shares.
Coursera Course Dataset & Recommendation System
Problem: Learners had no personalized way to navigate a large course catalog.
Approach: Recommendation system analyzing course ratings, difficulty, and enrollment data.
Result: Delivers personalized course suggestions.