Founding Lead Engineer - Agentic Development

The company

WORK-SELF builds Maya Enterprise, the Human Context infrastructure for enterprise AI. AI agents increasingly perform work inside large organisations, but they need context about the people and decisions around that work: who owns it, which decisions an agent can make, when a person reviews the result and how exceptions are escalated. Maya provides this governed context through an MCP-native server so employees and agents can complete work through accountable workflows.

We are a London-based enterprise AI company. In our first year we won enterprise contracts across banking, utilities, property services and European public finance, reached EBITDA-positive operations with a small team, and deployed Maya with enterprise customers. We work with technology and delivery partners and hold a granted US AI patent assigned to the company. Our next stage is scaling product delivery, enterprise adoption and partner-led growth. The people joining now will help build that next stage.

The role

You will lead the engineering of Maya's agentic runtime and Human Context infrastructure, taking features from design through production. Your work will connect structured employee context to enterprise AI agents while preserving permissions, review steps, escalation routes and auditability. This is a hands-on founding engineering role. You will write code, make architecture decisions, design for failure and build a dependable technical foundation with a small team.
Expected compensation ranges: £65k – £95k per year (Base) + £150k - £250k (Equity) + Bonus on performance and standard benefits

What you will own

• MCP-native Human Context infrastructure. Design and develop services, APIs and MCP interfaces that serve governed employee context to approved agents and enterprise workflows. Establish clear contracts for identity, ownership, decision rights and contextual access.

• Agent runtime and controlled autonomy. Build and improve runtime components for review, escalation, exception handling, human approvals, delegation and tool permissions. Define boundaries of agent autonomy, ownership and decision rights as explicit, testable contracts.

• Evaluations, traces and safe releases. Maintain traces tied to agent or tool versions, evaluation suites, shadow runs, regression checks and staged rollouts. Test thin or conflicting context, uncertain decisions, outages, failed tools and human takeover. Monitor correctness, latency, security and cost in production.

• Secure, versioned architecture. Shape data models, semantic retrieval, context access and integration boundaries with identity, tenancy, auditability and least-privilege access in mind. Treat model, tool and permission changes as changes that require controlled testing.

• Technical leadership. Write clear specifications, review code and architecture, mentor contributors, partner with product and implementation, and turn field learning into a prioritised engineering backlog.

You will know you are succeeding when

• First 30 days. You understand Maya's architecture, priority runtime flows and technical risks. You have started shipping features or integrations, including relevant partner-stack connectivity, using AI-assisted development responsibly, and have established a practical delivery and quality plan.

• By 90 days. You have shipped a substantial runtime capability or reliability improvement with automated tests, observable execution, documented failure handling and a clear human-review path. You use tools such as Claude Code, Codex, Vercel or Lovable where appropriate, and required security and compliance checks pass.

• By 180 days. You have made releases more reliable and shortened engineering delivery cycles, with measurable improvements in runtime quality, integration readiness and controlled agent behaviour. Security and compliance results are documented, and customer feedback demonstrates stronger product performance.

Who does well here

• Substantial hands-on software engineering experience, including systems you designed and shipped to production.

• Practical work on LLM applications, agents or orchestration, and confidence debugging model, tool and context failures.

• Strong APIs, backend services, data modelling and cloud engineering skills in languages such as Python or TypeScript, or equivalent.

• Strong grasp of MCP, identity, authorisation, secure agent tool access, retrieval, evaluation and production observability. Experience with versioned skills or agent-to-agent protocols is a plus.

• Ability to lead technical choices in a small team, explain trade-offs and deliver reliable work with limited specification.

This is a hands-on agentic engineering development role for someone who builds production systems, not a management-only position or a sequence of demos.

What you get

• Hands-on ownership of a foundational enterprise AI product and its architecture.

• Direct collaboration with the founder and short feedback loops from enterprise deployments.

• The opportunity to help build and scale an early-stage enterprise AI company.

If this resonates with you, we would love to hear from you. Any questions, please email us at careers@work-self.com