Data Science Engineer · 2025-Present
Savelife Ai
SaveLIFE Foundation, New Delhi
01
The Challenge
Over 150,000 people die on Indian roads annually. SaveLIFE Foundation monitors 200+ km of national highway corridors, but manually reviewing 500+ hours of dashcam footage was humanly impossible. Audit reports took weeks and stakeholder approvals stalled due to language barriers.
02
The Approach
Designed a multi-layered AI architecture combining local LLMs (for data sovereignty), RAG pipelines for context-aware violation classification, and MCP servers for orchestrating multi-model inference chains. All dashcam footage stays on Indian servers.
03
The Solution
- AI violation detection pipeline classifying Enforcement, Engineering, and Engagement gaps
- Geospatial telemetry engine: EXIF extraction + QGIS + Google My Maps integration
- GenAI Translation SaaS for 22+ Indian regional languages (React + Python + MongoDB)
- Full-stack dashboard on AWS with 99.9% uptime
- Local LLMs ensuring data sovereignty — zero footage leaves Indian infrastructure
04
The Impact
0+ Hours Processed
0% Detection Precision
0% Manual Effort Reduced
0x Faster Approvals
Tech Stack
Python React MongoDB AWS RAG LLMs QGIS Local LLMs
“When AI can process what humans cannot see at the scale humans cannot match, it stops being a technology story — it becomes a story about saving lives.”