About the Role
<p>For a young company in the rapidly growing AI implementation market, we are looking for a Tech Lead, starting as soon as possible, ideally in early November. The company runs LLM-based systems in production: content generation pipelines, retrieval-augmented generation (RAG) on internal documents, and automated workflows closely integrated with their business systems. The entire stack is built on TypeScript and Next.js.</p><p>These systems are already live. As Tech Lead, you take ownership of them, increase their reliability and quality, and expand them to new use cases, while being technically responsible for a team. The role brings together applied LLM engineering and solid backend engineering in TypeScript. Training or fine-tuning foundation models is not part of the job.</p><p>Stack: TypeScript, Next.js, Node.js, Postgres with pgvector, Docker, Azure, Anthropic and OpenAI APIs, Vercel AI SDK.</p><h2>What you’ll do</h2><ul><li><p>Technical responsibility for the team, including architecture decisions and code reviews</p></li><li><p>Technical guidance for the team and point of contact for non-technical colleagues</p></li><li><p>Take over and maintain the existing LLM pipelines: review the current architecture, identify failure modes, prioritise fixes, and refactor and extend the systems without disrupting production</p></li><li><p>Own the RAG systems from end to end: document ingestion and parsing, chunking, indexing, hybrid retrieval (BM25 and vector), query rewriting, reranking and grounded generation with citations</p></li><li><p>Implement and maintain chunk-level access control, index freshness and tenant isolation across all retrieval systems</p></li><li><p>Develop content generation pipelines that deliver consistent quality at scale, including human review steps</p></li><li><p>Build and run automated workflows against internal and third-party business systems (ERP, CRM, email, internal APIs), with durable and idempotent execution, retry and dead-letter handling, and approval steps for irreversible actions</p></li><li><p>Set up an evaluation framework for systems that currently run without one: golden datasets based on real production failures, retrieval metrics and more</p></li><li><p>Implement observability along the entire request path</p></li><li><p>Optimise cost and latency through prompt caching, batching, model routing and smaller models where it makes sense</p></li><li><p>Identify where deterministic logic is the better choice and implement it accordingly</p></li><li><p>Work directly with non-technical colleagues to define and validate automated processes</p></li></ul><h2>What you bring</h2><ul><li><p>Professional experience with at least one LLM-based system in production, including responsibility for its operation and incident handling</p></li><li><p>Advanced TypeScript and Node.js skills: strict typing of non-deterministic model output, async and concurrency patterns, streaming responses, structured error handling</p></li><li><p>Production experience with Next.js: App Router, route handlers, server actions, streaming to the client</p></li><li><p>Proven track record of taking over and improving existing codebases under production traffic</p></li><li><p>Hands-on retrieval expertise: hybrid search, choosing embedding models, cross-encoder reranking, metadata filtering, permission-aware retrieval and structured diagnosis of poor retrieval quality</p></li><li><p>Experience working with real-world documents: PDFs with tables, scanned material, DOCX, HTML, including layout-aware parsing, OCR and evidence-based chunking</p></li><li><p>Structured outputs and tool calling are part of your daily work: JSON Schema, Zod or similar runtime validation, function calling, handling malformed or partial output, context window management</p></li><li><p>Experience designing and running LLM evaluations</p></li><li><p>Experience with LLM tracing and evaluation tools in a TypeScript codebase (e.g. Braintrust, Langfuse, Promptfoo, OpenTelemetry or Arize Phoenix)</p></li><li><p>Familiarity with Postgres including vector search (pgvector or a comparable vector store), Docker, Git, CI/CD and at least one major cloud platform</p></li><li><p>Practical experience with the Anthropic and/or OpenAI TypeScript SDKs</p></li><li><p>Confident communication in English, German is a plus</p></li></ul><h2>Nice to have</h2><ul><li><p>Durable workflow execution for long-running, unattended processes (Temporal, Inngest or similar)</p></li><li><p>Agent orchestration in production, tool calling, recovery, multi-step workflows (Vercel AI SDK, LangGraph, Mastra, Claude Agent SDK, MCP TypeScript SDK)</p></li><li><p>Experience integrating enterprise systems such as ERP or CRM platforms like SAP</p></li><li><p>Security and data protection in LLM systems, defences against prompt injection and data exfiltration, PII handling, GDPR-compliant design, EU-hosted or self-hosted inference</p></li><li><p>Structured or graph-based retrieval for ent