→
I build end-to-end
Idea to live in days, not quarters. Product, backend, frontend, design — AI-native, not AI-flavored. I move across the whole stack so momentum never waits on a handoff.
Open to building together
Udit Khandelwal
Founder who ships end‑to‑end.
I build things and I never stop learning. Fifteen years in, a Principal Engineer — with time at Microsoft and Amazon — now co-founding Seerly and deliberately teaching myself the other half of the game: distribution. Great products don’t win on their own; someone has to get them in front of people. Lately, that someone is me.
15+years building
4companies, ex-Big Tech
10+products shipped
2ventures founded
01 — Who I am
A builder who never stops learning.
If I had to describe my style in one line: curious, hands-on, and allergic to standing still. I’ve spent 15 years widening my lane rather than guarding it — Android, distributed backends, cloud security, GenAI platforms, and now my own company. Principal Engineer at Microsoft (GenAI, multi-agent, RAG), SDE at Amazon (Alexa, Prime, Transportation), ingestion at Palo Alto Networks.
The constant isn’t a technology — it’s the loop: get curious, learn fast, build the thing, ship it, learn again. It’s why I can turn a weekend into a playable AR game, a stray idea into a trading system, or a new field into a real product. I’m an eternal learner by temperament, and right now the thing I’m learning hardest is the one most engineers avoid: how to sell and distribute what I build.
02 — How I operate
Build. Learn. Now — distribute.
∞
I learn relentlessly
New domain, new stack, new craft — I ramp fast and go deep. Fifteen years of that habit is why my range keeps widening instead of narrowing. Curiosity is the real engine.
☼
Now: cracking distribution
Building is only half the job. I’m deliberately moving into BD, sales and GTM — learning to get products in front of the people who need them. The engineer’s rigor, pointed at growth.
03 — Selected work
What I’m building & have built.
Flagship · live
↗
Seerly
AI visibility platform
As discovery moves from search to AI assistants, Seerly shows brands how they’re represented in AI answers, who gets cited beside them, and what content actually drives visibility and conversions — with proactive agents that nudge you on what to write next.
GEO / AI-SEOAgents0→1
Live
↗
SlayFit
AR webcam fitness games
No headset, no install — your webcam becomes the controller. Throw punches, dive to save goals, run in place to escape a dino. MediaPipe pose tracking + WebGL at 60fps, with browser-to-browser 2-player over WebRTC.
MediaPipePixiJS / WebGLWebRTC
Live · itch.io
↗
Relic Hunter
3D maze-runner
An endless temple-runner with a collapsing-floor mechanic, playable now on itch.io and headed to CrazyGames. Built the full loop — level ramp, audio, analytics, and a portal-ready launch plan.
3D / TIPGame designGTM
Open source
↗
MCP Bash Server
Dev tool · MCP
A local MCP server that lets Claude Desktop and Cowork run shell commands on your
machine — a run_bash tool over stdio or streamable-HTTP. (It’s the
tool that built and shipped this very site.)
Python · FastMCPstdio / HTTPMIT
Live · open source
↗
Review Assist
Dev tool · MCP
Review AI-written code as fast as agents write it. An MCP server turns a coding agent’s session into an Intent Document — what was asked, assumed, tried and tested — and a GitHub App renders it as a guided review on the PR at reviewassist.dev. Stores none of your code.
reviewassist.devMCP · GitHub AppApache 2.0
Open source
↗
Multimodal Agent Framework
Python agent library
A library for composing agents that play to the strengths of different LLMs — mixing OpenAI and Claude within one system. The scaffolding under a lot of my agent work.
PythonMulti-LLMMIT
Open source
↗
VNC AI Agent
Autonomous agent on a VM
A multi-user AI agent that runs on a remote VM and does tasks for you — driving a real desktop over VNC with clipboard sync, WebSocket streaming, and browser automation. Dockerized and deployed to the cloud, so the work happens without you having to do it.
VNCDocker + AzureBrowser automation
Ongoing
Quant & markets
Systematic trading, self-built
Designing an India-market trading and backtesting system — indicator research, Bank Nifty futures studies, and a research-analyst knowledge base. Engineering pointed at markets.
BacktestingSystematicData
2010 → today
Enterprise track record
The credibility spine
Microsoft (GenAI, Autogen, Copilot evangelism) · Amazon (Alexa Hints, Prime buying, Core Transportation) · Palo Alto Networks (Prisma Cloud ingestion) · Flipkart (Android) · NetApp · IBM.
Ex-Big TechPrincipal Engineer
04 — Writing & thought leadership
Writing about building AI, for real.
Active writer on Gen AI architecture, agentic systems, and the practical reality of shipping AI products. 13 articles on Medium plus regular long-form on LinkedIn. A few worth starting with:
📌 Pinned
Choosing Your AI Coding Strategy
After a year working alongside AI coding agents, the key to productivity isn’t just using them — it’s a practical framework for when, where, and how much to hand over. Lessons from hands-on experience.
Two-part deep dive
What’s missing in today’s AI app builders
A multi-part architectural deep dive into why the current generation of AI app builders falls short of what teams actually need — and what a more capable generation has to get right.
A Framework for Choosing the Right LLM Agent Architecture
Building a system that can autonomously create, monitor and heal applications surfaces real design hurdles. Here’s a framework for reasoning about which agent architecture actually fits the problem.
Builder’s Guide to Entrepreneurship
Lessons from 13 months after leaving a high-paying job — what I got wrong, what worked, and what I’d do differently. The honest version, for anyone weighing the same jump.
All 13 articles on Medium ↗ · long-form articles on LinkedIn ↗
05 — Let’s build & grow
Open to collaborations.
Founders, operators, engineers, and the merely curious — whether you want to build something in AI, consumer or dev tools, or you want to crack distribution and grow organic + AI-driven discovery with Seerly, I’d love to talk.