Repath
Khan.
Ray, to most people.
Founder, Chief Executive Officer
& Chief Engineer — Leemer Group
I run the company.
I also read the diffs.
I can’t leave things alone.
Most founder pages open with achievements. I’d rather open with a confession: I can’t leave things alone.
A form that asks for the same thing twice. A parliament whose record is technically public and practically invisible. An AI agent that announces it’s finished when it isn’t. I notice these the way some people notice a crooked picture frame. Then I can’t stop thinking about how it should have been built.
Leemer Group is what happens when you stop just noticing.
At the same desk.
Should this
exist?
Can this be
built well?
On paper I hold two roles. In practice they’re one question asked from two directions.
As Chief Executive, I ask: should this exist? Where we spend our time. Who we work with. What we’re willing to walk away from.
As Chief Engineer, I ask: can this be built well? Architecture, AI infrastructure, model evaluation, and the standard code has to meet before it ships.
When those questions are answered by different people in different rooms, companies tend to build things that are brilliant and unwanted, or wanted and fragile. I want them answered at the same desk.
Not every CEO should be an engineer. But Leemer is made of engineering, and I don’t want to lead it from a distance.
Mostly lessons. Some wins.
I didn’t start Leemer because I’d figured it out. I started it because figuring it out is the only work I want to do.
I’ve had businesses fail. I’ve had products die in a folder no one opened again. I’ve had ventures work better than I planned. Each one left a note in the margin.
Those years taught me the most useful thing I know: impressive and useful are different exams. Plenty of software passes the first and fails the second.
At Analog Devices I wrote software for a semiconductor manufacturing floor, where a bug isn’t a bad review — it’s a line that stops. You learn quickly that ‘works on my machine’ is not an engineering standard.
I studied Computer Science and left before the last chapter. Not because I’d stopped learning, but because I was learning faster outside the lecture hall than in it.
An experience log, not a repository history. Dates and venture names are deliberately unspecified. Tap an entry to read its margin note.
I studied Computer Science and left before the last chapter. Not because I’d stopped learning, but because I was learning faster outside the lecture hall than in it.
At Analog Devices I wrote software for a semiconductor manufacturing floor, where a bug isn’t a bad review — it’s a line that stops. You learn quickly that ‘works on my machine’ is not an engineering standard.
Then came years running operational businesses. Rotas, suppliers, cash flow, and customers who don’t care how clever your system is if the order is late.
Those years taught me the most useful thing I know: impressive and useful are different exams. Plenty of software passes the first and fails the second.
I’ve had businesses fail. I’ve had products die in a folder no one opened again. I’ve had ventures work better than I planned. Each one left a note in the margin.
I didn’t start Leemer because I’d figured it out. I started it because figuring it out is the only work I want to do.
Intelligence is
getting cheap.
Trust
isn’t.
Worth trusting.
Here’s what I believe about this moment.
Intelligence is getting cheap. Trust isn’t.
A model can produce an answer in a second; whether you can rely on it is a slower question. An agent can write a thousand lines before lunch; whether they’re correct is something it rarely checks. Public data can be open to everyone and still read by almost no one.
The distance between what technology can do and what people can safely rely on — that’s the gap Leemer works in.
We’re not here to put AI into everything. We want to know where it belongs, where plain deterministic software does the job better, and how to make the two keep each other honest.
The goal isn’t to make technology look intelligent. It’s to make it worth trusting.
Four doors. One sentence.
Leemer Labs — the workshop underneath everything. Research into models, inference and the infrastructure intelligent software runs on.
Critique.sh — the teammate a coding agent calls when it thinks it’s done. It reproduces, repairs and verifies, then hands back proof instead of promises.
Leemer Chat — advanced AI models in an interface built for doing work, not demos.
DáilDex — the Dáil, delivered to your inbox. Follow a TD and get plain-English email about what they actually did. Built on Oireachtas open data, free for citizens.
They are four products, each with its own identity. To me they’re one sentence: complex systems should get easier to use without getting harder to trust.
A developer shipping code at 2am and a voter wondering what their TD did this week have the same problem. The information exists. Making it useful is the hard part.
Leemer LabsOpen models, rigorous evaluation and practical AI infrastructure. A research lab built to turn experiments into things people can use.
Explore Leemer LabsOne group, four ways into the work. A group schematic, not a runtime dependency diagram.
Commitments. Not deadlines.
I don’t want Leemer to be the biggest collection of AI products. I want it known for thinking clearly and building carefully.
Research that becomes product. Product that exposes new research. Engineering that is ambitious without being careless. Independent thinking, whatever the market is doing that quarter.
Some of what we start will serve developers. Some will serve businesses, institutions, or people who’d never call themselves technical. Some of it won’t work.
That’s the price of building things that didn’t exist yesterday. I’d rather pay it than build only what’s safe.
- Think clearly.
- Build carefully.
- Keep learning.
- Stay independent.
I’m still building it.
Being a founder isn’t a title. It’s agreeing to be responsible for something that wouldn’t exist if you hadn’t insisted on it.
Responsible for the direction. For the calls nobody else wants to make. For the outage at 3am. And for whether the work deserved doing at all.
I still write code. I still pull apart systems I don’t understand. I still argue with my own technical decisions, and sometimes lose.
The day I’m too far from the work to understand it is the day I should step back and rethink how I lead.
I want Leemer to be a place where curiosity survives growth, engineers are respected rather than managed around, and the better idea wins no matter who had it.
That’s the company. I’m still building it.
Both lenses read the same here.