Projects
Things I've shipped
Three end-to-end builds — an ML prediction system on AWS, a GenAI content generator, and a cloud data pipeline. Each card links straight to the repo.
Flight Price Prediction — AWS SageMaker
End-to-end ML pipeline on AWS SageMaker: preprocessing, feature engineering from raw flight data (airline, route, timestamps, duration), and hyperparameter tuning with XGBoost via Bayesian optimization.
Engineered features using scikit-learn Pipelines, FeatureUnion, and custom transformers. Deployed a Streamlit web application to Streamlit Cloud for real-time price predictions.
LinkedIn Post Generator — Gen AI (Llama 3.2)
Two-stage GenAI pipeline: extracts topic, language, and length metadata from a user's past LinkedIn posts, then uses Llama 3.2 (via Groq API) to generate new posts matching their personal writing style.
Implemented few-shot prompting by retrieving stylistically similar past posts and deployed an interactive Streamlit application for on-brand post generation with minimal manual editing.
YouTube Trending Data Pipeline on AWS
Fully automated end-to-end data pipeline using AWS S3, Lambda, and Glue with the YouTube Data API to ingest, clean, transform, and store data following Medallion Architecture (Bronze/Silver/Gold).
Orchestrated with AWS Step Functions for reliable scheduling, retry logic, and automated alerts across pipeline stages ensuring consistent, analysis-ready data.