AI Engineer (Full-Stack)

• Build AI-powered web features and small products end to end: frontend, backend, database, deployment.
• Own the AI layer of our voice and text agent product: prompts, tool calling, speech-to-text and text-to-speech tuning, model choice, latency and cost per minute.
• Add AI features to our POS/ERP and 3D configuration products where they solve a real problem.
• Build evals and monitoring, so quality is measured, not guessed.
• Automate internal work for business development, marketing, QA and admin: drafts, reports, research, status updates.
• Work with QA on how to test AI output that is not the same twice.
• Send a short monthly brief to the team: what changed in AI, and what it means for our products.
• Recommend models and tools with a reason and a cost, and write the decision down.

• Two or more years of full-stack web development, with a modern frontend framework and a Node.js or Python backend.
• At least one LLM feature shipped to real users: tool calling, RAG or structured outputs, with any provider.
• You understand evals, token costs, latency, and what to do when the model is wrong.
• Working experience with a cloud platform and a database.
• You can explain a technical trade-off simply, to engineers and non-engineers.
• Clear written and spoken English.

• Voice AI: real-time agents, SIP, streaming speech-to-text and text-to-speech.
• Angular, NestJS or AWS Bedrock.
• Daily use of AI coding agents.
• Docker and CI pipelines.
• German or Arabic.

• The first AI engineer's seat, with the AI architecture yours to shape.
• AI features in front of paying users within weeks, in Germany and MENA.
• A CTO who builds with AI every day and will argue with your choices.
• Budget for models, tools and the time to try new ones properly.

You would report to the CTO and co-founder and work beside a QA engineer, a product designer, business development and marketing. You are the second builder on the team, so the work is real and the review is honest.

Four steps, usually inside three weeks. An informal call where you walk us through one AI feature you shipped. Then a short paid exercise in any stack you like: a small agent that answers from a menu, calls a tool to place an order, and comes with a few eval cases. A decision within a week of the last conversation.