The platform I lead is used by
I build AI products end to end. The architecture, the roadmap, the hiring, the incident calls and the cloud bill. Eleven years in software, seven of them leading teams, and still hands-on in the codebase.
2019 - Present
I lead engineering for an AI media platform used by Viacom18, Zee, Jio, Sony, Garena and BARC. Promoted from Senior Software Developer.
At Tessact I grew engineering from 2 to 10+ people without pausing enterprise delivery, and in earlier companies built teams of 15+. Technical hiring, onboarding, mentoring and performance conversations, across backend, frontend, AI and product design.
Monthly AWS bill
53%
lower, about $96K saved a year
Workload profiling, container right-sizing and CPU-only inference builds that dropped about 5 GB of unused CUDA from every image.
Designed the monetisation layer: subscriptions, usage-based credits, entitlement enforcement and billing. A services-led product became self-serve SaaS.
Transcoding, scene detection, VAD, transcription and LLM calls run on Celery and RabbitMQ workers, so the API stays fast under load.
PostgreSQL, Redis, Elasticsearch on Kubernetes
Find any moment across an archive, recut long-form for every platform, put an AI presenter on screen and turn footage into campaigns, all in one workspace.
Automated tagging, compliance detection, talking avatars and short-form video generation.
I also own production incident response for every major client account, run technical discovery, PoCs, SOWs and estimates, and set the engineering standard: architecture rules, pre-commit hooks and review gates.
Founder, UltraAer
Hear it once. Keep it forever.
A pendant that listens to the conversations that matter and keeps every decision, dose and deadline. Launching soon under UltraAer, my hardware company.
Prototype R6: three parts, two screws, no clips. Tray, deck and lid, with every component seated in its own bay. 26.6 × 53.7 × 15.3 mm.
The pendant streams audio over Bluetooth LE and the phone does the transcribing. No cloud speech-to-text, and every transcript is saved on the phone first.
Speakers, decisions, doses and deadlines are drawn out of the talk, in the languages people actually mix at home and at work.
Every conversation is kept and searchable. Answers come back with the moment they were said, and actions land before the meeting ends.
Outside Tessact I build hardware and software that runs at home, not in someone else's cloud.
Smart home control, made physical.
A privacy-first smart home platform running entirely on a Raspberry Pi 4B: a FastAPI backend, a native Qt 6/QML touchscreen kiosk, Home Assistant integration and an iOS client.
Offline voice pipeline: Porcupine wake word, faster-whisper STT, Llama intent classification and Magpie TTS, calling Home Assistant services locally.
Open source
Open-source Home Assistant kiosk and dashboard.
An Electron kiosk for Linux and Raspberry Pi touch panels with display rotation, power and brightness control, MQTT discovery and live system telemetry. Plus a responsive dashboard card that adapts from portrait wall panels to landscape displays.
Multi-arch releases (Debian arm64 and x64, plus portable builds) ship through GitHub Actions with a one-line installer.
Bengaluru
Promoted from Senior Software Developer
Bengaluru
Mumbai
Boston, MA (Remote)
New Delhi
Hiring an engineering leader, putting AI into a product, or curious about noti.? Send a note.
