Work with me
Three things, kept separate so you know which one you need.
In 2005 I had a working conversational AI system, a million dollars raised, and an investor who decided Google would get there first. He may have been right. What was missing in that room was anyone who had built something, lost it, and could say plainly what that was worth.
That is why the first of these three is free, and why it stays free. The longer version of the story is on the about page.
Founder mentoring
I work with three or four founders at a time. Not as a scarcity device — that is the number I can hold in my head well enough to be useful. If I am full I will say so, and you can ask again later.
What I am useful for: deciding what to build first and what to leave out, whether a technical claim someone made to you is real, hiring the first engineers, hardware — which very few people in this ecosystem have actually shipped — and recognising when a thing has stopped working. I have closed a company. I can talk about that part without flinching.
What I am not useful for: consumer growth marketing, anything that depends on a large advertising budget, and fundraising as a service. I will be honest about your deck. I will not carry it for you.
There is no fixed schedule. Most of it happens over a call when something is genuinely stuck, which is when the conversation is worth having.
AI and product advisory
Paid work goes through Neonflake, where I am a co-founder — about twenty engineers building AI and engineering products for clients.
Engagements are scoped and priced as fixed projects: a defined question, a defined output, a price agreed before anything starts. No hourly billing and no open-ended retainer, because both of those pay for time rather than for an answer.
- Whether an AI idea is buildable, what it would actually cost, and what the honest failure modes are
- Reviewing a product plan before it is committed to engineering
- Evaluating an LLM approach — fine-tuning, retrieval, or neither — where most of the real work turns out to be data quality and evaluation rather than model choice
- Technical due diligence on a company someone is about to buy into
If the work belongs with Neonflake's team rather than with me, I will say so and hand it over.
Speaking
Free for student events, universities and ecosystem programmes. Paid for corporate events. I travel within India.
Topics I can speak on properly, rather than ones that look good on a poster:
- Building conversational AI in India in 2005, and why it did not become a company
- Manufacturing hardware in India as a small company, including the parts that went wrong
- What separates AI projects that ship from AI projects that demo well
- Starting from nothing — in my case selling kites and tutti frutti before I was fifteen
How to start
Write to me. A few paragraphs is enough. Tell me what you are building, what is stuck, and what you want out of a conversation. If you are a company, tell me the question and the deadline. I do not need a deck.
I read everything. I reply to most of it.