1 day ago
Berlin, GermanySenior
Responsibilities
- Build and improve LLM agent workflows for nursing forms, assessments, care reports, and care-plan suggestions.
- Create production-trace and synthetic-scenario benchmarks, regression tests, and evaluations for ML and LLM systems.
- Investigate failures across the app, backend, and models and convert findings into fixes, tests, and data-labeling tasks.
- Train or adapt models when they improve quality, speed, cost, or privacy compared with frontier models.
- Optimize model routing, caching, latency, cost, and sensitive-data protection.
- Partner with product, engineering, data, and customer-facing teams and incorporate feedback from care professionals.
Requirements
- Experience training, evaluating, and shipping ML or NLP models to production.
- Experience building evaluations for ML or LLM systems and using error analysis and regression tests to guide product decisions.
- Hands-on production experience with LLMs, including agents or tool use, structured outputs, and context design.
- Strong software engineering skills with the ability to write tested, reviewable production code and learn unfamiliar languages or frameworks.
- Sound judgment with sensitive data, an end-to-end approach to ambiguous problems, and ownership in a fast-paced product environment.
- Fluent English.
Benefits
- Work on a product with direct feedback and clear social impact.
- Opportunity to shape the agent stack and grow ownership of a domain, market, or technical area.
- Collaborate with engineering and product teams on training, evaluation, and production ML.
- Collaborative environment emphasizing written communication, shared learning, and regular time together.
- Stock options and access to language courses.
Tech Stack
Categories
About voize
voize builds a voice-powered AI assistant that lets nurses and caregivers document care at the bedside via smartphone, with integrations to existing care IT systems. It sells its software to care providers and focuses on reducing administrative workload and improving documentation quality; the company is YC-backed, founded in 2020, and headquartered in Berlin. The product is used in 600+ care facilities in German-speaking markets and is built to be GDPR-compliant.