Agentic Software Engineer

London · Full-time · On-site

AI is collapsing the time and effort required to turn ideas into working software. We're redesigning how we build around that — using agents across the software lifecycle and enabling more of our team to ship changes directly.

We're hiring an experienced software engineer who is already pushing beyond task-by-task use of agents. You'll engineer the systems around agents that let them take on increasingly substantial work — running for longer, working in parallel and involving you when your judgement is actually required.

This isn't a model-training role. We're focused on what happens around the models — engineering the systems that turn increasingly capable agents into a fundamentally different way of building software.

Hult is a global business school that teaches a Computer Science for Business degree. The engineering team doesn't sit adjacent to that mission — it's part of it. How we build software, adopt new tools and think about automation feeds back into what we teach.

We give engineers real ownership. You'll pick up open-ended problems, shape the approach and have the backing of a team that trusts you to land them.

We ship fast and iterate constantly. We want the distance between an idea and something running in production to be as short as possible.

AI has already changed who can ship software here. You'll help us push that further — giving more of the team the tools, context and guardrails to turn ideas into production changes without engineering becoming the bottleneck.

We're serious about discovering what AI-native engineering looks like in practice. We don't have all the answers, and part of this role is finding them.

Own outcomes, not tickets. Work directly with product owners and stakeholders to understand problems and find the shortest responsible path from idea to production. You'll use engineering judgement — and agents — to close the gap between request and delivery.

Delegate outcomes, not steps. Your goal is to be able to say: "Here's the outcome, constraints and evidence I expect. Go progress this work and involve me when my judgement is actually required." You'll design workflows where agents can plan, execute and verify work rather than waiting for you to tell them what to do next.

Engineer the agent harness. Build the instructions, skills, tools, permissions, environments, validation and feedback loops that allow agents to reliably complete substantial engineering work. You'll understand these as engineering primitives rather than a fixed recipe, and continually experiment with how they fit together. When an agent fails or requires intervention, you'll ask what could change in the harness to prevent it next time.

Engineer context and memory. Design how agents discover and retain what they need to know about our systems — architecture, conventions, decisions, product intent, operational state and previous work. Give agents the right context at the right time without simply giving them more context.

Push toward an autonomous software factory. Help work move from intent through planning, implementation, verification, deployment and observation with progressively less synchronous human intervention. Enable multiple agent workstreams to run in parallel and converge on tested, shippable outcomes.

Make autonomous work trustworthy. Generating software is increasingly cheap; knowing whether it's good is harder. Build tests, evaluations, observability and feedback systems that establish whether work is complete without requiring a human to inspect everything the agent produced.

Raise the capability of the whole team. AI has already enabled more of our team to ship changes. Turn successful approaches into reusable capabilities that allow people to safely take increasingly ambitious ideas into production.

• Strong software engineering foundations. You have substantial production software engineering experience and the judgement to understand architecture, production systems and risk.
• You've materially changed how you work because of AI. You regularly delegate substantial engineering work and spend more of your time defining outcomes, providing context and designing verification than directing implementation step-by-step.
• You systematically increase the scope of delegation. When an agent needs you to tell it what to do next, you ask whether better context, tools, verification or structure could have allowed it to progress independently.
• You think in systems, not prompts. You think about the environment an agent operates in, how it gets what it needs, how it progresses work and how you know when it has succeeded.
• You care deeply about verification. You don't trust generated output because it looks plausible. You build ways to establish correctness without relying on reading everything yourself.
• You can operate from an outcome. You're comfortable working directly with technical and non-technical stakeholders, understanding what matters and finding a pragmatic path to production.

Experience with particular languages or technologies matters much less to us than the ability to understand unfamiliar systems quickly.

Experience building AI systems can also be highly relevant where the underlying problems transfer to agentic software engineering.

We want to see evidence of how you actually work.

Your application will ask you three short questions about your experience with agents: something you've built recently, how you've increased what you can delegate, and something you've tried that didn't work.

We expect you may use AI to help with your application — we use it constantly too. That's fine. What we're looking for is your experience and your thinking. Specifics matter much more than polished writing, and we'll use your answers as the starting point for the interview.

And don't feel constrained by our questions. If there's something that better demonstrates how you work — a project, repo, harness, experiment, write-up, demo, or anything else you think we'd find interesting — show us. We'd much rather see something real than read another paragraph about how passionate you are about AI.

This is a full-time, on-site role based at our Chelsea office in London, reporting to the Engineering Manager.

We're trying to discover what software engineering looks like when implementation is no longer the primary constraint.

If you're already experimenting at that boundary, we'd like to talk.