About us
Kiso Health is developing an AI-powered therapy platform for individuals facing complex psychiatric disorders. Our product is an MDR-certified medical software (smartphone app) that combines evidence-based psychotherapy with cutting-edge artificial intelligence.
Our mission: to improve outcomes for severe mental health conditions and provide long-term digital support through intelligent, empathetic, and safe AI.
Aufgaben
Your Role
- Design and integrate AI components for empathetic user interaction and personalized therapy
- Build a Reinforcement Learning system (e.g., Contextual Multi-Armed Bandits)
- Implement a modular LLM architecture with prompt routing, fallback strategies, and safety layers
- Develop and integrate explainability features (e.g., SHAP, confidence scores, audit trails)
- Collaborate closely with psychologists, UX specialists, and product teams
- Ensure compliance with MDR, GDPR, and the EU AI Act
Qualifikation
Your Profile
- 3+ years of experience in Machine Learning / Reinforcement Learning / LLM applications
- Familiarity with small datasets and online learning techniques (e.g., bandit algorithms)
- Strong skills in TensorFlow, PyTorch, or JAX
- Experience with LLM API integration (e.g., GPT, Claude), prompt engineering, and safety layer design
- Proficient in explainable AI, logging, GDPR compliance – ideally in a healthcare context
- A passion for building AI that is empathetic, secure, and truly helpful – for people who need real support
Benefits
What We Offer
- Full ownership of the AI domain – from strategic planning to technical execution
- Opportunity to build a meaningful AI system with real medical and social impact
- Equity through VSOP – share in the success of our mission
- A highly interdisciplinary team with clinical, psychological, and technical expertise
- Remote-friendly setup, flexible working hours, and fair compensation
Ready to help us rethink therapy?
We’d love to see a project where you:
- Used a bandit system or RL model for personalization
- Developed an LLM system with routing or safety layers
- Built explainability features for sensitive applications
We look forward to hearing from you!