UIDAI Intelligent Policy Dashboard
UIDAI Hackathon • January 2026
Description & Objective
Objective: To build an intelligent dashboard that accelerates data processing for UIDAI policy governance, eliminating manual analysis through ML-powered insights.
Description: Implemented a Velocity Governance ETL pipeline with automated data cleaning, parsing 232,000+ unified logs into an intelligent dashboard. Used K-Means clustering visualized on maps and linear regression forecasting to generate Priority Intervention Lists for high-risk migration zones.
Technology Stack
- Python
- Scikit-Learn
- Dash
- K-Means Clustering
- Linear Regression
Key Achievements
- Accelerated data processing, saving 30 hours/week with automated ETL pipeline.
- Eliminated manual demographic analysis through ML-based clustering and forecasting.
- Parsed and processed 232,000+ unified log records.
Technical Highlights
- ETL Pipeline: Velocity Governance with automated data cleaning
- ML Models: K-Means clustering, Linear regression
- Visualization: Interactive maps, Priority Intervention Lists
Resource Links
Drishti — AI-Powered Crowd Safety App
GDG Hackathon • December 2025
Description & Objective
Objective: To create a real-time crowd safety monitoring system using computer vision and AI for anomaly detection in crowded environments.
Description: Implemented a client-side computer vision pipeline utilizing OpenCV for video frame extraction and MediaPipe for real-time crowd density counting (MAPE under 20%) and bottleneck prediction. Designed a multi-modal situational awareness agent that processes critical frames through Google Gemini 1.5 Flash for anomaly detection, with Firebase for authentication and event logging.
Technology Stack
- OpenCV
- MediaPipe
- Firebase
- Google Gemini 1.5 Flash
- Python
Key Achievements
- Real-time crowd density counting with MAPE under 20%.
- Multi-modal AI agent for anomaly detection using Gemini 1.5 Flash.
- Secure user authentication and event data logging via Firebase.
Technical Highlights
- CV Pipeline: OpenCV frame extraction + MediaPipe
- AI Model: Google Gemini 1.5 Flash for anomaly detection
- Backend: Firebase for auth and data logging
Resource Links
Safety First Tourism App
Smart India Hackathon 2025 • September – October 2025
Description & Objective
Objective: To develop a full-stack tracking platform for safe tourism with real-time monitoring, geo-fencing, and dynamic safety scoring.
Description: Built a full-stack tracking platform featuring reliable anomaly flagging, real-time traveler monitoring via OpenStreetMap, and geo-fencing to secure user journeys through dynamic safety scores. Engineered a risk-aware itinerary planner with a self-curated SQLite database of 250 mapped Indian locations for secure and customized travel routing.
Technology Stack
- React
- Express.js
- OpenStreetMap API
- SQLite
- JavaScript
Key Achievements
- Real-time traveler monitoring with anomaly flagging and geo-fencing.
- Dynamic safety scores for secure user journeys.
- Self-curated database of 250 mapped Indian locations.
Technical Highlights
- Frontend: React with OpenStreetMap integration
- Backend: Express.js with SQLite database
- Features: Geo-fencing, risk-aware itinerary planner
Resource Links
Intelligent QnA System
Bajaj HackRx 6.0 • June – July 2025
Description & Objective
Objective: To build a Retrieval-Augmented Generation (RAG) pipeline for processing and answering questions from unstructured insurance documents.
Description: Engineered a FastAPI RAG pipeline integrating vector chunking and Named Entity Recognition (NER) to process and index unstructured insurance documents up to 30 pages long. Reduced natural language query response times by 40% by architecting an efficient FAISS vector search system, using the Gemini API to synthesize top-k retrieved chunks into precise, context-aware answers.
Technology Stack
- Python
- FastAPI
- Gemini API
- FAISS
- NER
- RAG Pipeline
Key Achievements
- Reduced query response times by 40% with efficient FAISS vector search.
- Processes unstructured insurance documents up to 30 pages long.
- Context-aware answer synthesis using Gemini API.
Technical Highlights
- Backend: FastAPI with RAG pipeline architecture
- Vector Search: FAISS for efficient retrieval
- NLP: Named Entity Recognition + Gemini API
Resource Links