Building theimpossible.zero to 1.from scratch.breakthrough.
Founder-engineer with 5 successful acquisitions. I transform bold ideas into products that scale.
Track Record
Companies Built
First hire → Acquired by Airtable
Early engineer → Acquired by Alteryx
QuickOffice acquisition
View all acquisitions
Acquired by Broadcom
Founding engineer → PubMatic
A clear pattern over time
I prefer hard starts and real users. Across my career I have joined early, shipped the product, helped set a clear story, and supported teams until the work was stable. Several companies I worked on later got acquired. That is not luck. It is a way of working.
Dopt
Founding Principal Engineer
Built a developer onboarding platform with SDKs, components, visual flows, and AI guidance. Full stack across Node.js, TypeScript, Fastify, gRPC, Prisma, PostgreSQL. Front end in React with react flow, react aria, vanilla extract. Infra on GCP with Docker, GKE, CloudSQL, Turborepo.
Trifacta
Principal Engineer
Helped the shift from on prem to SaaS with a product led motion. Redesigned core user workflows like the visual flow and transformation experience. Started and grew the frontend group in India to more than 50 engineers. Platform powered Google Cloud Dataprep.
MobiPrimo
Founding Engineer
Early work on backup, sync, and sharing across mobile and desktop.
What zero to one feels like
Three short stories
The router
Day one can be basic. I connect the Wi Fi router. New repo. First commit. CI is up. Environments named. We move from idea to action.
The glue
We have a prototype. I listen to the first ten users. I trim scope. I make docs and a starter kit. Sales and support get a clear message. We become easier to try and easier to buy.
The shape
Now we scale what works. We add seams in the right places. Owners are clear. We automate painful steps. We measure what matters. The product becomes a habit for the customer.
Featured work
Key projects that went from zero to acquisition
Dopt
Founding Principal Engineer
Developer onboarding with SDKs, components, visual flows, and AI guidance
Full stack architecture and build. Front end system. Workflow editor. SDK design. CI and CD. Developer experience. Work with design and GTM on story and launch.
Trifacta
Principal Engineer
On prem to SaaS with a product led motion
Redesigned core workflows. Performance work with React and TypeScript and D3. Started and scaled a 50 plus engineer frontend group in India.
Google Chromium Office
UI Tech Lead
Office document viewing and editing inside Chrome
Plugin design and UI. Decisions from usage and crash analytics. Accessibility work.
VMware NSX
Senior Member of Technical Staff
Clear UX for large scale network security
Angular, D3, EXT JS. Complex data visuals. Production delivery.
How I work
Principles that guide every project, from day zero to successful exit.
On Day 0, I've been the person who connects the Wi-Fi router and mops up the build pipeline. And on Day 700, I've been the person who says "we can simplify this system and still ship by Friday."
Ship early and often
Stay close to users
Small sharp teams
Clear story before scale
Make the best path the fastest path
Measure and remove what does not help
Building delivXchange as an AI native platform
I am in build mode on delivXchange. It is AI native from the start. We use agents, voice, MCP, RAG, n8n workflows, and agentic deployment. This is how it fits together and why it matters.
Input Channels
Multi-channel customer intake for orders and support
Voice agent
Real time speech in and speech out with streaming. The voice agent takes calls for new orders, order edits, status checks, and FAQs.
Social and messaging
We watch DMs and messages through connectors. A social intake agent turns a message into a clean order or support ticket.
Intelligence Layer
AI agents that understand context and plan actions
Agents as the core
We model work as tasks that a small set of agents can own. Each agent has a clear goal like intake, ordering, routing, or support.
RAG
We keep menus, modifiers, prices, hours, and store policies in a vector store. Agents ground responses in that data.
Agentic planning
For multi-step tasks, the agent writes a short plan, runs one step, checks the result, then continues.
Processing
Smart routing and tool integration via MCP
AI inference
We route work by cost and need. Small models handle classification. Larger models write replies or plans when complex.
MCP tools
We use MCP to plug tools and data into agents. POS, delivery partners, maps, payments, and CRM appear as standard tools.
Execution
Automated workflows that integrate with POS systems
POS workflows
Each order flows through n8n recipes that validate items, apply rules, write to POS, and confirm back to the user.
Stack
TypeScript and Node for services. React for UI. WebSockets for live updates. Postgres as source of truth.
Quality & Safety
Continuous evaluation and safe deployment practices
Evals
We keep gold tasks from real calls. Before changes, we run offline evals and compare success, correctness, tone, and cost.
Deployments
Each agent ships behind flags. We can turn on features for one store, one region, or a small percent.
Safety
We monitor token use, latency, and error rates. We set guardrails for PII, tool scope, max spend, and rate limits.
Value
The goal is simple. Fast intake from phone, web, and social. Clean routing to POS and delivery. Less back and forth for staff and customers. A system that learns from real work and gets a little better every week.