Projects
A collection of open-source tools, experiments, and production systems I've built or contributed to.
RAGineer
Production-ready local RAG Text-to-SQL chatbot that converts natural language into safe PostgreSQL queries with 84% SQL accuracy.
- 84% SQL accuracy, 80% end-to-end, ~2.5s latency
- SQL injection prevention, rate limiting, 64 tests at 97% coverage
- Multi-model: Qwen2.5-Coder (84%) vs CodeLlama (78%)
Real-Time Fraud Detection
High-performance FastAPI microservice achieving 90% latency reduction (50ms → 5.4ms) with full production observability.
- False Positive precision from 6% → 78%, 83% Recall maintained
- MLflow experiment tracking, class-weighted XGBoost
- 100% observability: Prometheus + Grafana dashboards
GetAnime
RAG-powered semantic anime recommendation engine with sub-second latency and 95% relevance accuracy across 12,000+ entries.
- Sub-second query latency with Groq LLM + ChromaDB
- 99.5% uptime on GCP with Kubernetes + Grafana monitoring
- Automated ETL with HuggingFace Sentence Transformers
Mental Health Predictor
AI-driven tool predicting depression among students using TensorFlow, trained on 27,000 records with LIME interpretability.
- Neural network on 27,000 student records
- LIME interpretability for transparent predictions
- Personalized insights and self-care recommendations
Data Analysis Toolkit
Interactive toolkit for EDA, feature engineering, model building, and visualization with Plotly and scikit-learn.
- End-to-end: cleaning → feature engineering → model training
- Interactive Plotly visualizations
- Download modified datasets
E2E Text Summarization
Production-ready NLP summarization pipeline with CI/CD, Docker, and AWS EC2/ECR deployment.
- CI/CD with GitHub Actions, Dockerized deployment
- AWS EC2 hosting + ECR image registry
- ROUGE and SacreBLEU evaluation metrics
Heart Disease Prediction
ML-based heart disease risk assessment comparing Logistic Regression, Random Forest, Gradient Boosting, and SVM.
- 85.2% test accuracy with Logistic Regression
- Interactive Streamlit web application
- Comprehensive model performance visualization
Note Taking App
Full-stack Django note management app with authentication, rich text editing, search, and collaboration tools.
- User authentication and note CRUD
- Search, filtering, and file uploads
- Rich text editor with collaboration features