HI, I'M
HEERA SHANKER

A Retrieval-Augmented Generation pipeline integrating RxNorm, OpenFDA, and RxNav APIs for real-time drug interaction and contraindication alerts. Built a multi-stage architecture — entity extraction → normalization → parallel async retrieval → LLM synthesis — using Llama 3 with deterministic outputs via a Streamlit UI.

Binary classification pipeline on 10,000+ customer records. Evaluated Logistic Regression, Random Forest, SVM, and XGBoost. Designed 12+ engineered features via EDA with Pandas & Matplotlib. Deployed a real-time Streamlit dashboard for churn probability predictions.

Trained regression models (Linear, Ridge, Lasso, Gradient Boosting) on the Kaggle Ames dataset (1,460 samples, 81 features). Engineered 20+ features with missing-value imputation, categorical encoding, and outlier handling. Deployed an interactive Streamlit app.

A student performance analytics dashboard. See the repository list on GitHub for the code and current status.

Full-featured AI Voice Assistant built in Python with speech recognition (SpeechRecognition), text-to-speech (pyttsx3), and multi-command NLP processing. Integrated OpenWeatherMap, Wikipedia, Gmail SMTP, YouTube, and Chrome automation. Modular architecture with CPU monitoring, screenshot capture, and music playback.

Recommends products from text descriptions using natural language processing techniques. Built end-to-end in Python with NLP preprocessing, similarity scoring, and ranking.