Agentic Software Engineer
- Compensation
- Not disclosed
- Location
- London, London, United Kingdom
- Work setup
- Onsite
- Seniority
- Not specified
- Employment
- Full time
- Experience
- Not specified
- Category
- Software engineering
- Posted
Skills
Must have
- 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 th
- 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 progr
- 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.
Full description
Agentic Software Engineer
London · Full-time · On-site
TL;DR
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.
Why Hult?
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.
What you'll do
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.