End-to-end data science systems, analytics, and ML pipelines
A selection of practical, production-minded ML and analytics work with clear outcomes and thoughtful delivery.
Shipped end-to-end ML apps with containerization and cloud-ready runtime patterns.
Automated testing, reproducible training pipelines, and stable model artifact workflows.
Built retrieval-augmented systems for grounded question answering in regulated domains.
Three representative builds focused on predictive modeling, pricing analytics, and GenAI retrieval systems.
Retention Analytics
Flask-based churn prediction app with decision recommendations and ROI-style retention guidance.
Identify high-risk customers early enough to trigger targeted retention action.
scikit-learn pipeline with model metadata tracking, probability thresholds, and deterministic action decisioning.
Deployed web app, API endpoints, Docker support, and CI test workflow.
Regression Analytics
FastAPI app for insurance charge estimation with reproducible training and JSON/API usage.
Estimate expected medical insurance charges from customer attributes.
Regression pipeline with ingestion, transformation, model training, and model-card style documentation.
FastAPI UI + JSON endpoint, Docker packaging, and CI checks.
LLM + Retrieval
Modular Retrieval-Augmented Generation pipeline for banking documents with strict answer guardrails.
Answer policy and product questions using only trusted internal banking documents.
PDF ingestion, chunking, embedding, Chroma vector store, retriever filters, and guardrailed prompting.
FastAPI endpoint + Gradio chat UI, Dockerized demo mode, readiness checks, and evaluation harness.
Portfolio
Core repositories I currently maintain and deploy.
Churn prediction with retention decisioning and ROI guidance.
Insurance charge estimation with API-ready regression outputs.
Grounded Q&A assistant for banking documents and policies.
See more on GitHub.
I can walk through architecture choices, metrics, and deployment workflows.