I build products from zero to production. I was the founding engineer at a climate-tech startup, and at Awiros, which runs video AI for airports and smart cities, I design, build and ship products across the whole stack: the UX, web and mobile apps, services, data pipelines and the platform that installs them on customer sites. I like owning the whole loop, from the first Figma frame to fixing things on site. SDE II since June 2026; company-wide Best Performer two cycles running.
UVIP product · UXAwiros's video-intelligence platform. I led its ground-up rebuild in React 19 and TypeScript and owned the product and UX. 31 modules · 300+ typed APIs on one live cache · I wrote 318 of its 372 test files |
CV app delivery platformHow 50 computer-vision apps reach customer sites. I re-architected it as versioned per-app Docker images that Kubernetes Jobs pull and install on site. Replaced a 3-hour monolithic Jenkins build · deploys 90%+ faster |
Developer Mode developer experienceFor computer-vision developers: mount an app's source onto a live pipeline and watch its alerts in real time. No release cycle between a code change and a real result |
Seek searchNatural-language person search across video alerts. I designed and built it in Go on Milvus, gRPC and Kafka. Used at 33 Indian airports · 650K alerts a day · median search 51 s → 4.7 s |
Enterprise SSO securityLDAP/Active Directory and OIDC presets for Keycloak, Entra ID, Okta and Google Workspace, with just-in-time provisioning and group-to-role mapping. Shipped to customers |
Delhi IGI Airport operations end to endThe backend behind the airport's operations dashboards, which I built and ran in production on site. 9,189 e-challans filed with Delhi Traffic Police · trolley planning per baggage belt from live AODB arrivals · a LightGBM forecasting service |
Two mobile apps React NativeBuilt solo in Expo: GMR's airport operations app, and Aura Companion with alerts, live activity and an AI chat. Expo, RTK Query and EAS builds, from the first screen to release |
Video summarisation and crowd estimation productShipped for Delhi Safe City. Crowd analytics built for 2,500-camera sites on pre-aggregated ClickHouse tables |
Event pipeline dataA ClickHouse pipeline I architected, handling 20M+ events a day. Bangalore Safe City (3,500+ cameras): downtime −99%, storage 5 TB → 450 GB |
Vision-model playground productBoth ends, NestJS API and React UI, of a catalog where users browse and compare computer-vision models. Built end to end |
Background-job engine reliabilityBulk imports rebuilt as resumable background jobs with crash recovery, auto-retry and 60 s stall detection, plus report-kit, a shared PDF engine. Removed the record cap on reports |
Live configuration and service split platformPer-site JSON config (hand-edited, restart to apply) became live settings managed from the UI, and the Node.js monolith became independently scaling API and Kafka-consumer services. Configuration changes without restarts |
This work lives in private company repositories. More detail on LinkedIn.
Enveave founding engineer · 2022–24Co-led the climate-tech startup's 0-to-1 MVP with the founders: architected 6 core features and designed 15 Figma prototypes. AWS costs −33% · delivery time −50% with CI/CD |
Tortoise 2024Lessor and seller dashboards for an electronics rental platform with 10,000+ customers. Admin load time −40% with RTK Query caching |
cse.iiitd.ac.in webThe official IIIT-Delhi CSE department website, built in Next.js with five other students. Content lives in Google Sheets, so staff update it without code |
Sahaayak bioinformaticsSequence-alignment and ORF tools for biology students, from my Computer Science and Biosciences degree. Next.js + Flask |
| Product and frontend | Backend and data | Infra and delivery |
|---|---|---|
|
|
|
|
My work commits live in my employer's private repositories. A private repo here replays one timestamp per real commit, with no code or messages, so the graph matches my actual activity.



