Web-based SaaS application for organizing & analyzing automation tasks. Integrated Groq API for AI-powered summaries and recommendations, with Google OAuth, session management, and CSV/JSON/PDF output generation.

Built end-to-end LLM-powered platform that extracts delivery commitments from Slack and Gmail, scores slip risk 0–100, and auto-alerts owners before deadlines are missed.

Developed machine learning workflows for combustion instability prediction using TCN models and Bayesian Optimization. 8-channel real-time sensor fusion achieving 92% accuracy, outperforming LSTM baselines.

Cardiovascular digital twin combining ML risk prediction, explainability, physiological data acquisition, and real-world data processing. Awarded Novel Contribution Award at National Conference.

End-to-end hyperspectral image processing pipeline for automated crop health assessment. Built an ML ensemble using Random Forest and SVM, achieving 89% disease prediction accuracy.

Deep learning time-series classification workflow for real-world applications with severe class imbalance. Data preprocessing, feature engineering, and model evaluation.

Engineered CNN-based image recognition system achieving 94% accuracy in waste classification across 6+ categories with automated preprocessing pipelines.

AI-powered Android productivity application mapping natural language task prompts into Google Calendar based on daily predicted energy levels.