SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: EUR 65,000 - 80,000 / annual
cleverly.de is building an AI-powered learning companion for children, serving over 60,000 students through 1:1 mentoring, tutoring, and live courses. This is a three-year research and development project funded by the Berlin Investment Bank (IBB) under the ProFIT Berlin program, offering rare opportunity to build with proper time and budget rather than shipping features on a sprint cycle.
You will design and build the LLM architecture powering the learning companion: retrieval over child-specific learning profiles and course content, proactive communication triggers, and real-time generation of micro-learning materials. You'll work closely with the CTO, frontend developers (React/Node.js and .NET/Angular), Head of Product, and pedagogical leadership.
Key responsibilities:
- Design and build LLM architecture including orchestration, RAG over learning profiles and course materials, proactive triggers, and structured outputs (quizzes, fill-in-the-blank, multiple choice)
- Build context pipeline: prepare and vectorize heterogeneous data (course materials, interaction data, coach feedback, conversation histories) in Postgres/pgvector
- Implement evaluation and quality assurance: build eval frameworks for groundedness, age-appropriate language, and pedagogical appropriateness, including guardrails and escalation paths to human coaches
- Ensure privacy by design: GDPR-compliant processing of sensitive minor data, select and operate models on EU infrastructure and German data centers
- Integrate AI components into existing platform via REST APIs, working with the development team
- Evaluate technologies systematically: assess models and frameworks (LangGraph, Microsoft Agent Framework, DSPy), favoring modern proven tools over fine-tuning base models
The role offers technical decision-making authority: you build the AI layer from scratch with control over architecture, tool selection, and evaluation strategy. You'll see measurable impact—60,000 children using a product families actually rely on.
Requirements:
- Completed degree in Computer Science, Data Science, or equivalent qualification
- 3–5+ years professional backend development experience with Python, Node.js, or C#
- Production experience with LLM applications: RAG pipelines (embedding selection, chunking, hybrid search, re-ranking), agent orchestration, and/or prompt and pipeline optimization
- Experience with LLM evaluation: offline eval sets, metrics for factuality and relevance, guardrails (stated as the single most important criterion—ability to demonstrate system improvement rather than claim it)
- Solid SQL and relational database knowledge; experience with vector databases (ideally pgvector) and non-relational databases
- Structured, documented working style and enthusiasm for small high-performing teams
- Very good German or English proficiency (significant content and collaboration is German-language)
- Genuine enthusiasm for the vision and mission of children's education
Nice to have:
- Experience with self-hosted model inference (vLLM, TGI) and Linux server administration
- Basic understanding of Node.js or C#/.NET to anticipate integration points
- Experience with regulatory requirements, especially GDPR and processing minor data
- Affinity for education and EdTech
- A/B testing and event-based analytics experience
