Das ist der Job
The aim is a product that feels noticeably better than what people are used to today.
Darum lohnt es sich
About the Business Our client is a growing technology business developing software that applies modern AI models to everyday tasks. It's a small, senior team by design, with fast, shared decision-making and a strong emphasis on getting things right.
The engineering work is demanding: outputs need to hold up over longer interactions, stay grounded in context, and keep working reliably even though the underlying models don't always behave predictably. The Role In this role, you'll take AI capability and turn it into something that actually works for end users.
You'll take ownership of problems from start to finish – shaping how the model behaves, building what sits around it, and making sure it holds up once it's live. It's a role that spans machine learning, engineering, and product, with one underlying aim: getting AI to perform well in everyday use, not only in controlled demonstrations.
What You'll Be Doing Taking AI-driven features from concept through to a shipped, working product Designing and refining the prompts, tools, memory, and workflows that drive agent behaviour Converting raw model output into something structured, dependable, and predictable Tracing and fixing problems anywhere in the stack - model, orchestration layer, infrastructure, or interface Tuning for speed, cost, and dependable performance in live usePutting together simple, practical evaluation methods that reflect real usage Working closely with product and engineering colleagues to turn loosely defined problems into working solutions Signs of Success Live models consistently hit the accuracy, speed, and reliability bar expected of them Problems in production get spotted early, diagnosed properly, and fixed at the source The pipelines, training routines, and inference systems behind the product stay dependable and easy to maintain Strong working relationships across engineering, product, and research help features land reliably Changes to the models and systems are grounded in real usage data and lead to measurable gains Tools
