Building AI-Powered Products with Engineering Precision.

End-to-end mobile solution for farmers to manage crops and access guidance. Built with a scalable Flutter architecture and Firebase backend, enabling real-time data sync, authentication, and extensible modules for future ML/IoT integrations.

Full-stack telemedicine application enabling secure patient–doctor communication. Implemented user authentication, appointment scheduling, and role-based access control with a focus on clean REST API design and relational data modelling.

AI-powered document assistant that lets users query documents conversationally. Implemented text chunking, embedding generation, and retrieval workflows, then wired them into an LLM pipeline to return context-aware answers.

Research-grade framework for linking user identities across platforms using cross-modal embeddings. Designed and evaluated a pipeline that combines profile text and interaction signals to compute similarity, targeting robust identity resolution across noisy data.

Suicide risk detection pipeline. Engineered an asynchronous prediction API wrapping a fine-tuned Roberta model. Decoupled inference service from the React client using a message queue pattern to handle burst traffic reliable.

Developed a system to classify emotional states (happy, sad, angry, neutral) from audio signals using MFCC and spectral feature extraction. Evaluated SVM, Random Forest, and KNN classifiers across multiple datasets (RAVDESS, CREMA-D, SAVEE, EMO-DB) to analyze generalization and feature robustness.