The business
We are two founders with three exits between us. We have raised more than $15m from world class investors, and our companies have generated more than $50m across those exits. We build consumer products, and this one is funded from our own profits. There is no VC timeline, no fundraise we need to close, and no plan to raise external capital, so we can build for the long term without the usual pressure to keep financing the business. Salary is top of market and stable, with a defined path as the business grows.
The product
You step into a book as a character. You decide what you do and say, and the story moves around you. The people in it remember what happened, want things of their own, and do not always do what you expect.
Millions of people already spend hours a day talking to fictional characters, but the products they use are chat windows with a costume on. Nothing remembers properly, nothing builds, and nothing pays off. A story here may run for hundreds of turns. Different characters have to remember different things, relationships have to change, old events have to matter again at the right moment, and the story has to stay surprising without drifting. All of that has to disappear behind an experience that feels like reading. Nobody has done it well yet, because it is a hard AI problem and a hard product problem at the same time.
The role
You are the first engineer. You work directly with the founders and the founding product designer to build the product, take it to market, and improve it quickly from what real readers do. You will ship the whole product, from the database to the screen the reader sees, and you will use coding agents to do most of the typing. Your time goes on the things agents are still bad at: knowing what should be built, designing the experiment that tells us whether it works, and deciding where AI can and cannot be trusted to make the call.
What you will own.
– The production loop. Data model, streaming, background work, concurrency, payments and the credit ledger, deployment, observability. Every turn must be reconstructable after the fact, because failures in a system like this are hard to reproduce.
– Benchmarks and evals. Test cases for continuity, character behaviour, narrative quality and failure modes, developed alongside the first working loop so they tell us what to build rather than only whether we built it.
– Memory, context and models. How a story is remembered across hundreds of turns, which models do which jobs, retrieval, orchestration, cost and latency. We have starting points. We expect you to make the simple approaches fail on our own test cases before we build anything bespoke.
– The standards the next hire inherits. Enough structure around agents that speed does not turn into unmaintainable code.
A world class former CTO and two senior technical advisers are on call as a sounding board. The founders come from product and growth, have worked closely with engineers before, and know how to give you what you need and then get out of the way.
How we work
The bar is a tier A consumer product: story quality, latency, typography, state and reliability all count, and we will push on all of them.
We would rather test three approaches in a week than debate them for a month. We expect you to push back when we are wrong, and we will say so plainly when we think you are. You are never stuck alone: founders are available daily, advisers are a message away, and you will have the tooling, compute and model budget to run real experiments. If something is slowing you down, we remove it.
What we are looking for
– AI first. Models, context, memory, evals and coding agents are part of how you naturally think about software, and you know where they fail.
– A real product engineer. You have shipped consumer products end to end, not worked only in notebooks or research settings.
– You have built and operated a production system with language models at its core, seen it fail with real users, understood why, and been responsible for fixing it.
– You design the experiment before you reach for custom infrastructure, and you can build an eval that measures something rather than producing a reassuring score.
– Strong on databases, APIs, concurrency, background work and production reliability. Agents can write a migration; you have to know whether it can destroy fifty thousand running stories.
– Design instinct, and the confidence to make the call when nobody hands you a spec.– Genuine interest in why people come back to a product and what makes something feel good.
– You can explain a hard technical decision in plain English, and you are comfortable challenging founders and advisers.
– You want unusually broad ownership and a path to Head of Engineering as the team grows.
Requirements
Non negotiable: real production engineering depth, and the ability to build with models, context, memory and evals as a natural part of how you work.
Preferred: several years as a full stack product engineer, experience as one of the first engineers at an early stage company, and a strong technical education such as IIT, NIT, IIIT or BITS.
Stack
Web first. We currently expect TypeScript, Next.js and Postgres, frontier models over API, and background jobs for anything that should not sit on the live reader path. None of it is sacred. You will have real say in what we use, what we buy, what we borrow from existing AI systems, and what is worth building ourselves. If you would choose differently, tell us why.
_1548133325694.png){kind=logo}
