Building end-to-end AI-powered systems — from QR-based provenance platforms to deployed ML pipelines and smart robotic prototypes. I turn data into decisions and ideas into production software.
I'm a B.Tech student in AI & Data Science at REVA University, maintaining a 9.18 CGPA while shipping production-grade software. My projects span the full stack — from React frontends and FastAPI backends to ML pipelines and IoT robotics.
What drives me is the gap between data and decisions. Whether it's a QR-based agricultural provenance system that cuts verification time by 80%, a churn model deployed with a live API, or a geolocation-powered places recommender — I build things that solve real problems and run in production.
When I'm not coding, I'm volunteering at university fests, competing in hackathons (Top-10 twice), and pursuing certifications from IBM, Tata, and Datacom.
QR-based agricultural provenance system enabling real-time verification of produce — reducing verification time by ~80%. A 3-tier web app handling batch registration and retrieval in under 1 second using JSONBin cloud storage and qrcode.react.
End-to-end deployed ML system on 7,043 records. Logistic Regression in a Scikit-learn Pipeline (OHE + StandardScaler) achieves ~80% accuracy with zero data leakage. FastAPI on Render exposes a public /predict endpoint; Streamlit Cloud provides the interactive frontend.
Context-aware recommendation app delivering real-time, mood-driven destination suggestions. Fetches up to 50 POIs per query via Geoapify Places API with sub-10-meter precision. Achieved a 95+ Lighthouse score. Zero-framework, component-based UI with dark mode and favourites.
5-table normalised MySQL schema (1NF–3NF) covering students, courses, enrolments, grades, and instructors with full foreign-key integrity. VIEWs for reusable grade reports and TRIGGERs for automated integrity checks. Complex JOINs and aggregations power top-performer analytics.
Production-ready Streamlit app with a zero-code Auto-ML pipeline and a natural language "Smart Assistant" for rapid dataset profiling and automated anomaly detection. Transforms raw CSV data into interactive Plotly visualisations and executive PDF reports — reducing manual analysis time by ~60%.
Intelligent robotic prototype using Embedded C for real-time obstacle detection with 90% navigation accuracy. Automated path-correction routines decreased route latency by 20% — translating core autonomous vehicle concepts into a functional, low-latency embedded system.