AI Applications Architect (LLMs, TypeScript, Node.js)

An established software product company is putting AI at the centre of what it builds, and it is looking for one engineer to own that end to end. This is a hands-on architect seat: you take the AI stack from prototype through to production, building on top of foundation models rather than training them, and you ship the features that prove it works.

You will not be writing speculative documents. The direction you set is direction you have already proven in code — spikes, prototypes and reference implementations the rest of the engineering team can build on. Expect your week to run between shipping production TypeScript, designing agents and retrieval, and coaching the developers who will own this alongside you.

What you’ll do

• Take end-to-end ownership of the product AI stack, from prototype through to production.
• Design, build and ship AI-powered features and services — agents, retrieval, tool-use, orchestration and evaluation — as working software, not just designs.
• Own model selection, prompt and agent design, context management, cost and latency trade-offs, and the evaluation harness that keeps quality measurable.
• Stand up the AI platform primitives: model gateways, tracing and observability, evaluation and regression suites, and safe rollout mechanisms.
• Establish the patterns, guardrails and reusable building blocks that let the wider team add AI capabilities safely and consistently.
• Deliver production features across the full stack, TypeScript and JavaScript front to back, integrating AI capabilities into the platform.
• Write, review and merge production code regularly — setting the engineering bar through the code you ship, not only the standards you write.
• Build robust APIs and services that expose AI capabilities to product teams with clear contracts, versioning and backwards compatibility.
• Engineer performance, security, testability and maintainability in from the start, including privacy-aware handling of personal data.
• Uplift the engineers around you through pairing, code review, working examples and hands-on mentorship, and give practical technical coaching to team leads and developers across the C#, Angular and TypeScript stacks.
• Keep architecture just enough — the written records and diagrams that earn their keep, always backed by something running.
• Evaluate emerging AI tooling and models, run focused proofs of value, and make clear, cost-aware recommendations.

What you’ll need

• An IT qualification, or extensive hands-on technical development experience.
• A proven track record shipping production software as a senior or lead full-stack engineer — code and delivered products you can point to.
• Deep, current expertise in JavaScript and TypeScript across the stack: Node.js server-side and a modern front-end framework.
• Hands-on experience building on top of LLMs and foundation models — prompt engineering, retrieval-augmented generation, tool-use and function calling, agents, and evaluation. Training or fine-tuning models is not the focus of this role.
• A solid understanding of the SDLC, API design, testing, debugging and production support.
• An understanding of scalability, high availability, application performance monitoring and cloud platforms.
• Strong application and data security knowledge, including careful handling of personal data and awareness of privacy regulation such as POPIA and GDPR.
• The communication to model complex ideas simply, and to present and sell an architectural design to the people who have to back it.
• Facilitation, training and leadership instincts — you pull others into the design and research phases to build buy-in and share what you know.
• Sharp analytical and critical thinking, a real appetite for learning, and the discipline to work to a deadline both independently and inside a self-managed team.

Nice to have

• Angular experience — a strong plus given the front-end stack here.
• Experience with the Anthropic stack: the Claude API, and agentic tooling such as the Claude Agent SDK or Managed Agents.
• Familiarity with .NET and C#, SQL Server, and integration patterns with observability and OpenTelemetry tooling.

What’s on offer

• The AI stack is genuinely yours to own — model choice, patterns, guardrails and platform primitives, prototype through to production.
• An architect seat that still ships. This is a builder’s role, judged on working software rather than documents.
• The chance to grow AI engineering capability right across an organisation, so the stack ends up owned by the team and not by one person.
• Room to assess emerging models and tooling properly, with focused proofs of value instead of guesswork.

Cape Town. Hybrid. Permanent.