I build data products and AI-powered analytics tools that turn complex datasets into clear business insights.
- 🤖 Building GenAI & NLP systems — trade intent detection, sentiment analysis, and LLM-powered analytics tools
- 🧾 Developed AI Receipt Tracker — production-style system using LLM vision to parse receipts, categorize spend, and generate analytics dashboards
- 🎬 Built recommender systems — content-based (TF-IDF + cosine similarity) and collaborative filtering (MovieLens)
- 💳 Built fraud detection pipelines using Random Forest & XGBoost with revenue optimization analysis
- 🛒 Performed customer segmentation on 500K+ retail transactions using RFM modeling, K-Means clustering, and behavioral analytics
- 📊 Strong in tree-based ML models — Decision Trees, Random Forest, XGBoost with hyperparameter tuning and cross-validation
- 📈 Experienced in ML evaluation — RMSE, MAE, Precision@K, Recall@K, clustering validation & dimensionality reduction (PCA)
Currently exploring LLMs, RAG pipelines, and AI-powered data workflows.
🏅 IBM Certified Data Scientist & Data Analyst
🎓 Machine Learning Specialization — Stanford
☕ Powered by curiosity, coffee, and F1 🏎️
A production-grade AI-powered receipt parsing and spending tracker built end-to-end
| Feature | Details |
|---|---|
| 🧠 AI Parsing | GPT-4.1-mini vision + GPT-4o-mini text — reads PDFs and images |
| 🗄️ Database | PostgreSQL (Supabase) + SQLAlchemy ORM with migrations |
| ☁️ Cloud Storage | AWS S3 for receipt file storage with presigned URLs |
| 🔁 Retry Logic | Tenacity retry wrapper for rate limits & API timeouts |
| 🧹 Smart Categorization | 20+ categories with sub-categories + category memory from history |
| 🔍 Duplicate Detection | SHA-256 file hashing to prevent double uploads |
| 📊 Analytics | Plotly dashboards — spend by category, store, month + drill-down |
| 🔐 Auth | Password-protected with Streamlit session state |
| 💰 Cost Tracking | Per-call token logging with monthly API cost monitoring |
| 🧾 Multi-store Support | Handles Costco, Target, Walmart — store-specific tax & discount logic |
Tech: Python • Streamlit • OpenAI API • PostgreSQL • SQLAlchemy • AWS S3 • Plotly • pandas • boto3
| Project | Description | Stack |
|---|---|---|
| 🎬 CineMatch | Content-based & collaborative filtering recommender with full ML evaluation | Python, TF-IDF, Cosine Similarity, KMeans |
💳 Credit Card Fraud Detection  |
Fraud detection with revenue optimization | Python, Random Forest, XGBoost, SHAP |
| 🚢 Titanic Survival Prediction | Feature engineering + 6 model comparison achieving ~87% accuracy | Decision Tree, XGBoost, SVM, KNN, Random Forest |
| 📈 Trade Intent NLP | Buy/Sell intent detection using POS tagging & sentiment analysis | Python, NLP, ML |
| 🏎️ F1 Sentiment Analysis | Sentiment analysis on F1 tweets | Python, NLP, Twitter API |
| 🛒 E-commerce Customer Segmentation | Customer & product analytics on 500K+ UK retail transactions — K-Means clustering, RFM analysis, revenue seasonality, cancellation trends | Python, K-Means, Seaborn, Pandas |
Languages
ML & Data
NLP & Gen AI
Cloud & Data Engineering
Tools
Visualization & BI
