Full Stack Agentic Developer

At Umain, we're not just creating software; we're crafting the future, one innovative project at a time. With our motto «Shape, Ship, Scale,» we empower our team to not only envision the future but also play a pivotal role in building it. We're a dynamic team where you can shape the future, ship groundbreaking solutions, and scale your abilities in many different ways.

We’re looking for a full stack developer who already works agentically. You write clear specs, run multiple agents in parallel against them, choose the right model for each job, and take full ownership of what ships.

You’ll work on client products end to end, from shaping requirements with stakeholders through to production. You’ll also help define how Umain builds software with agents: the practices, tooling and standards the rest of the team uses.

How we build

We start from a clear spec, break the work into tasks agents can complete, run those agents in parallel, verify the output with tests, evals and review, then ship. The spec is the source of truth. Agents do much of the implementation. You own the spec, the orchestration and the quality of the result.

Spec driven delivery

• Define requirements in a structured way: identify the actors, describe their use cases, then derive the features that serve them.
• Turn client needs into specs that both people and agents can build against: clear scope, acceptance criteria, interfaces and constraints.
• Keep specs, tests and code in step as the product changes.
• Break work into units that agents can complete and verify independently.

Agent orchestration

• Run and coordinate agent swarms: parallel agents, subagents and specialised roles (planning, implementation, review, testing).
• Build and maintain the supporting tooling: MCP servers, agent skills, prompts, context files and isolated workspaces.
• Design the checks that keep agent output trustworthy: test coverage, evals, guardrails and human review points.

Model selection and mixing

• Combine frontier and open-weight models to deliver results.
• Route work by capability, cost, latency and data sensitivity, including self-hosted or private deployments where client data requires it.
• Measure model performance on real tasks and change the mix when the evidence says so.

Full stack engineering

• Build secure, scalable web applications with TypeScript, React and Next.js.
• Build backend services and APIs with Node.js and NestJS.
• Create reusable UI components using design system principles.
• Work with cloud infrastructure (GCP, AWS or Azure), infrastructure as code with Terraform, and CI/CD pipelines.
• Build observable systems with metrics, logging and tracing, including for LLM and agent behaviour in production.

Clients and team

• Work directly with clients and cross-functional teams across design, product and data.
• Explain agentic delivery, its trade-offs and its limits to technical and non-technical audiences.
• Share what works: pair with colleagues, document patterns and help raise the agentic practice across Eidra.

• 3+ years of professional full stack experience, with strong TypeScript, React/Next.js, Node.js and core web fundamentals.
• Hands-on experience delivering production software with coding agents (for example Claude Code, Codex, Cursor or similar), beyond autocomplete.
• Strong requirement definition experience, working from actors to use cases to feature definitions with clear acceptance criteria.
• A working spec driven development practice. You can show specs you’ve written and what agents built from them.
• Experience running multiple agents in parallel and coordinating their output.
• Practical experience using both frontier and open-weight models, with a view on when each is the right choice.
• Strong engineering judgement. You read and review agent output critically and you own the result as your own code.
• Comfort with ambiguity, and the consulting instinct to work it through with clients.
• Curiosity. The tools change monthly and you enjoy keeping up.
• Building agentic features into client products: tool calling, RAG, MCP servers, multi-agent workflows.
• Running open-weight models in production (for example vLLM, Ollama or managed inference).
• LLM evaluation and observability tooling.
• Contributions to open source agent tooling or published writing on agentic engineering.
• Tailwind CSS, Storybook and design system work.
• Experience in a consultancy or agency setting.

Don’t match every point? Apply anyway. We care more about how you think and what you’ve built than about ticking every box.