Senior Software Engineer, Agentic Delivery

GCP and Claude Code

Team: Global Product Engineering

Reports to: Global Chief Product Officer

Partners with: Global Head of Engineering (Architecture)

Location: Remote. United States or South Africa (Cape Town hub).

Type: Full-time

Why this role exists

Incubeta is the leading AI Outcomes Agency. Our Seamless Suite turns proven engineering patterns into client products on Google Cloud, and our delivery model is agentic: small pairings of domain experts and engineers shipping production features in days, not quarters. We are looking for a senior engineer who lives in Claude Code and knows GCP cold, someone who can take that model and accelerate it hard.

This is a force-multiplier seat, the strong technical second on delivery. You will not own architecture governance. That sits with our Global Head of Engineering, who defines the target architecture and the standards our work conforms to. Your job is velocity and shipped outcomes within it: convert intent into working software fast, raise the quality bar through automation, and share the patterns and tooling that make an already strong team move faster.

What you will do

• Build production features on the Seamless Suite with Claude Code as your primary environment, running autonomous build, test, and iterate loops rather than hand-typing every line.
• Pair with practice leads and subject-matter experts to turn their intent into shipped features on a weekly cadence.
• Stand up and harden services on GCP: BigQuery data layers, Cloud Run deployments, Vertex AI and Gemini integrations, IAM least-privilege, Terraform, and CI/CD.
• Design and run agent workflows and MCP servers that automate the parts of the SDLC that should not need a human, including test generation, security scanning, documentation, and release checks.
• Drive feature velocity you can measure, including PRs shipped, cycle time, defect rate, and adoption, and report it cleanly so the investment is visible.
• Set the standard for what good agentic engineering looks like, and coach existing engineers and contractors into the same way of working.

A typical week

Our pods work in a fast loop: brainstorm and scope early in the week, build and test mid-week with Claude Code doing the heavy lifting, then validate and ship by Friday. You will run several of these loops in parallel, each one a small product unit, with the architect reviewing against the target architecture before merge.

What you bring

Must have

• Senior or staff-level software engineering experience, with a track record of shipping production systems other engineers depend on.
• Deep, hands-on GCP expertise: BigQuery, Cloud Run, IAM, Terraform, and CI/CD on Google Cloud. You can design, deploy, and secure a service end to end.
• Fluency in agentic coding with Claude Code or an equivalent, including running autonomous loops, writing effective specs and context, and orchestrating agents rather than only prompting them.
• Strong Python and TypeScript, with clean, typed, async-aware code and solid testing instincts.
• Comfort with production hardening: observability, secrets management, least-privilege access, and safe release practice.
• Self-direction. You can take an ambiguous brief and a domain expert and return working software, then explain the tradeoffs.

Strongly preferred

• Vertex AI, Gemini, and the Agent Development Kit (ADK); experience building or hosting MCP servers.
• Data engineering depth across pipelines, transformations, and warehouse modeling, given how much of our value sits in the data layer.
• React or a modern front-end framework for the product UI layer.
• Experience in product SaaS or a fast agency or consulting environment serving enterprise clients.

Where you can stand out

We will happily hire a strong generalist who nails the must-haves. You will stand out, and accelerate us, if you bring real depth in any of the following, because they map directly to where the product is going. You do not need all of them. Pick two or three you are genuinely strong in.

• Production RAG, done properly. Private, in-client-cloud retrieval (a glass-box model), hybrid and semantic search, reranking, chunking strategy, and retrieval evals. Not a notebook demo, a system that holds up on messy enterprise data.
• Knowledge graphs and GraphRAG. Building and querying knowledge graphs, entity resolution, and graph-backed reasoning that lets an agent answer what is connected to what, not just what is similar.
• Generative creative at scale. Image and video generation (Imagen, Veo, and the wider toolkit), brand-safe asset variance, and pipelines that turn a single brief into hundreds of on-brand executions.
• Multi-agent orchestration. Agent swarms and orchestrator and sub-agent patterns (ADK or equivalent), where specialized agents hand off, review, and run in the background rather than one model working alone in a chat window.
• Evals and guardrails for non-deterministic systems. LLM-as-judge harnesses, regression suites, tracing, and cost and latency control, so agentic features are measurable and safe to ship.
• Agent memory and context engineering. Durable memory, context-window strategy, and structured tool-use and function-calling design that keeps agents reliable across long, real tasks.
• Multimodal and voice. Agents that reason across text, image, document, and audio, not text alone.
• Synthetic data and model adaptation. Generating training and test data, light fine-tuning or adapter work, and a considered embeddings strategy.

How we work and what we offer

• Agentic-first engineering. We invest in the tooling, including Claude Code, API compute, and CI/CD, so you spend your time on judgment rather than boilerplate.
• High autonomy and high trust, with a clear architecture path so speed never costs us stability.
• A direct line to product leadership and real influence over how a global agency rebuilds itself around AI.
• Remote-first within the United States or South Africa, working with a distributed European and global team.
• Compensation: [senior or staff band, to be set with People for the relevant market].

How this role fits

You partner closely with our Global Head of Engineering, who owns architecture and standards, and you report to the Global Chief Product Officer. The boundary is simple: the architect owns how we build safely at the platform level, and you own how fast we turn ideas into shipped, client-ready features within that frame.