IIT Delhi graduate and Technology & AI Lead at the Bharti Institute of Public Policy, ISB Hyderabad. I build production-grade AI systems across computer vision, voice AI, geospatial intelligence, and retrieval-augmented architectures.
I take ML from research to deployment, alongside applied research in GNSS positioning and AI for public policy.
Four production systems, built end to end.
Geospatial intelligence, computer vision, agentic AI, and retrieval — each taken from data capture through models and serving to the interface on top.
Field geodata capture · spatial modelling
Geospatial data platform: an offline-first Android capture app and a web platform, running the inventory and boundary mapping pipelines.
stack
Monocular metrology · instance segmentation
Metric tree diameter from a single RGB frame — trunk segmentation plus pinhole projection, no reference object or specialised hardware.
stack
Multi-agent system · cross-sector analytics
One analytical system for agriculture, nutrition, rural development, and livelihood programmes — explored through four AI assistants for policy, maps, data, and reports.
stack
Retrieval · multi-agent generation
Citation-grounded RAG over 102 chapters of published research — page-level sourced answers and long-form briefs checked by an independent reviewer agent.
stack
Including Van Saathi, BarkVisionAI, and ArthaSakhi, plus case studies and the interactive project network.
Built for everyone, free to use.
Two browser-first platforms I design and run outside my research work — no installs, no accounts, open in a tab and start.
runs in your browser · nothing uploaded
Small tools. Serious work.
Free browser tools for developers and geospatial analysts — JSON, hashing, GeoJSON and EXIF, image annotation, charts, PDFs, and local AI for OCR and transcription. Files are processed in your browser; nothing is uploaded or stored.
scene 02 / 05
Presentations that turn.
“Make scene 2 a map”
share a link · export pdf / pptx
Presentations that turn. Not slides that sit.
Turns ideas, datasets, and research into animated presentations that play in the browser. Pick one of 20 story types, shape it by chatting (“make scene 2 a map”), then share a link or export to PDF and PowerPoint — and generate a promo video from your own slides.
Research and open data, published.
156,001
annotated bark images
A large annotated bark-image benchmark that lets models identify tree species from bark alone — no leaves, flowers, or fruit needed in the field.
5.2 GB
open field imagery
Single RGB images for estimating tree diameter at breast height, with a hand on the stem as scale — no stereo or LiDAR.
285
phone models · 15 makers
Raw GNSS captures from Android phones in real-world conditions, measuring how positioning quality varies across devices.
talks & workshops
FLARE Conference 2024 · University of Notre Dame
OGC Stack Winter School 2024 · IIIT Hyderabad
I’m always up for conversations with people working on vision, geospatial, or agent systems — research collaborations, product builds, or a problem you’re stuck on.
$ whoami --status
$ connect --list
$