Real AI Systems I’ve Built — Not Just Models
I build production-ready AI solutions: ML pipelines, time-series deep learning, GenAI (RAG), FastAPI services, and dashboards — designed for real business outcomes.
Featured Projects
These are end-to-end builds — data → modeling → APIs → UI → deployment. Each one is designed to be demo-ready and client-ready.
Built an end-to-end predictive maintenance system for Remaining Useful Life (RUL) estimation using turbofan engine sensor data, with a deployed API + dashboard and a GenAI copilot for explainable insights.
- Benchmarked ML baselines (Random Forest, XGBoost) then trained LSTM/GRU/CNN-LSTM/Hybrid models.
- Best performance reported on FD003 using a tuned GRU (RMSE ≈ 12.8, MAE ≈ 9.1).
- Unified inference pipeline + Dockerized services deployed on Google Cloud Run.
- Integrated RAG (FAISS + LLM) for contextual maintenance explanations.
Built a churn prediction pipeline with multiple ML models, class imbalance handling, evaluation metrics, and a Streamlit app for real-time churn scoring.
- Trained Logistic Regression, Random Forest, XGBoost, and an ANN model.
- Used SMOTE + scaling + hyperparameter tuning (GridSearchCV).
- Evaluated using AUC, Recall, F1-score, and confusion matrix.
Researched and evaluated SDS for feature selection across multiple real-world datasets, comparing results against common baselines.
- Achieved ~80–95% feature reduction with ~5–15% accuracy improvement over baselines.
- Compared SDS against RFE, PCA, and Mutual Information.
- Benchmarked across several ML models (SVM, XGBoost, Random Forest, Logistic Regression, KNN).
Want a Similar System Built for Your Business?
If you need an AI chatbot, predictive analytics dashboard, or automation workflow — I can build it end-to-end and deploy it.