Founding AI Engineer at Healf

[https://app.ashbyhq.com/api/images/user-content/b5f7de27-4bd6-4d48-b6ed-ec89eb566843/7f6782d8-2749-472d-a68e-38abdcdf80a8/1750099265864.jpeg] HEALF IS EUROPE'S FASTEST-GROWING COMPANY. Number one on the FT1000, number one on the Sifted 100. From £1m to over £100m in under three years, with a small, talent-dense team and an electric culture with day one founder intensity. Now we're aiming for £1bn in the next three. We curate the world's best wellbeing brands across The Four Pillars™: EAT, MOVE, MIND, SLEEP. That's the first chapter. THE NEXT CHAPTER IS HARDER AND MORE INTERESTING. We are moving from one market to many, from e-commerce to a technology platform, and from curating wellbeing to defining it. We are a health company, so we think we should act like one. At its fullest expression, Healf redefines what wellbeing means for tens of millions of people. WHY THIS ROLE IS HEALF We think the companies that win the next five years are the ones that put a capable AI agent behind every function in the business, faster and more reliably than anyone else. Operations, supply chain, customer experience, marketing, finance, compliance. Work that used to take a team now takes an agent and a person who knows how to direct it. This is not a research role and it is not a conventional engineering role. AI engineering is the functional label. The real job is taking a problem from any corner of the company and turning it into a production agent that quietly does the work, week after week, without breaking. The best version of this person has strong judgment about how to build. They do not reach for the heavyweight tool every time. When an agent is a short build, they use what is already there, Claude's managed agents or something off the shelf, and they ship it the same week. When the problem is genuinely complex, they reach for LangChain, LangGraph, Pydantic AI, and instrument it properly in LangSmith. They know the difference, and they never over-build. Somewhere right now there is an engineer who has built agents that real teams depend on. Not demos. Systems that take action, recover from failure, and earn trust. They are not on the market. But they are past the point where their current work still teaches them something, and they are ready to build the next thing. This is the role that person will take. WHAT YOU'LL OWN The agents. You will build production agents across every function at Healf. You will be hands-on in the code, shipping, not directing from a distance. The work is broad by design: operations and supply chain one month, customer experience or marketing the next. The build approach. You will own how we build agents. Off-the-shelf and managed agents where the job is small and speed matters. Frameworks like LangChain, LangGraph and Pydantic AI where the problem demands it, with evaluation and observability through tools like LangSmith built in from the start. Choosing well, every time, is the job. So is knowing when not to build at all. Reusable building blocks. You will drive velocity by building reusable building blocks. Every agent should leave behind components the next one can use, so the team gets faster with each build instead of starting from scratch each time. The unhappy path. Anyone can demo an agent that works. You build for the times it doesn't. You design for failure, retries, edge cases and clean handoffs to a person when the agent should not decide alone. You ship, watch how it behaves with real users, and tighten it. An agent is finished when it can be trusted on its worst day, not its best. Build versus buy. Frameworks, managed agents, models, infra. You will own these calls with authority and the context to make them well, and you will not default to building when buying is faster and just as good. The engineering principles. Not a style guide. The actual principles for how we build agents that can be trusted. The core one is evaluation: you decide what a good answer looks like before you ship, build a way to measure it, and keep measuring so quality never quietly slips when a model or a prompt changes. On top of that, where a human stays in the loop, and how we trace what an agent did versus what it claimed. You will define what good looks like here. The frontier. The tooling changes every few weeks. You will own staying ahead of it. You will research what is new in agentic AI, test what is worth testing, and bring the best of it into how the team builds. Then you will level up the people around you, so Healf is always building with the best available approach, not last year's. The technical direction. You will report to the Head of Data and AI and work hand in hand with the CTO. This is not advisory. You will have real authority over how AI gets built across the company. A small team, over time. You will start as a builder and grow a tight group of engineers around you as the work scales. This role is hands-on first and leading second, in that order. WHAT YOU'LL DELIVER IN THE FIRST 12 MONTHS - A growing library of live agents across multiple functions, each removing real hours of manual work every week - A growing set of reusable building blocks, so each new agent ships faster than the last - Measurable revenue and cost impact attributable to agents you built - A fast, repeatable path from business problem to deployed agent, measured in days, not quarters - Evaluation and observability built in, so agents are trusted rather than hoped for - The early nucleus of an AI engineering team built around your judgment WHY YOU'RE HEALF You have built agents or automation that run in production and that people rely on. Not experiments, not notebooks. Shipped, live, used. You have strong build versus buy instincts. You have used managed and off-the-shelf agents to ship something in a day, and you have built something far more involved with a framework when it was worth it. You can tell which a problem needs, fast. You build for the unhappy path. You think about how an agent fails before you think about how it demos, and you measure whether it actually worked once it's live. You have built something from zero and watched it reach real usage. You know both directions, the early ambiguity and the later scale. You go deep. When someone pushes on how a system actually works, under the abstraction, you go further in, not quieter. You stay at the frontier. You track what is new in agentic AI, form a fast view on what is real and what is hype, and fold the good parts into your work and the team's. You are close to the business, but this is a building role first. You spend most of your time in the code, and you bring stakeholders towards you with working agents rather than being pulled into endless alignment. You measure agents by the work they remove and the value they create, not by how clever they are. You are not looking for the next job. You are looking for the thing worth building next. THE REALITY This is hard. It requires speed and precision. It requires shipping autonomous systems into a live business, where a bad decision has real consequences, and doing it without the luxury of consensus. Your agents will sometimes be wrong. What matters is that you build the evaluation, observability and guardrails so the errors get caught, get smaller, and never repeat. If you want to spend the next two years building something foundational that the rest of the company will build on for a decade, this is that opportunity. SHOW US - An agent or automation you took to production. What was the problem, what did you build, and what work did it remove or what value did it create? - A build versus buy decision with real stakes. When did you reach for something off the shelf, and when did you go to a framework? What did you choose, why, and what happened? - How you knew an agent was actually good. How did you measure quality, and how did you catch it slipping when something changed? - An agent you made trustworthy. How did you evaluate it, observe it, and keep a human in the loop where it mattered? - Something new you adopted before it was obvious. A tool, model or technique you brought in early. How did you decide it was worth it, and what did it change? - A moment you saw how something would break six months before it did. What did you do with that knowledge? - Evidence you can work across the business. Not just clean systems. Agents that serve real teams and real users. THE DEAL Competitive and meaningful compensation package for the right person. We ask a great deal of the people who work here. We expect full ownership and a genuine commitment to give this chapter everything you have. In return, we will give you the same: everything we have, invested in your growth, your wellbeing, and the defining skills of the next decade. We have built the fastest-growing company in Europe with a team small enough that every person in it shapes the outcome. That is still true today. The next person we hire will change the trajectory of the company. IF THE MOST IMPORTANT WORK OF YOUR CAREER IS AHEAD OF YOU, THIS IS THE PLACE TO DO IT.