Motivation

Lifestyle-related diseases account for up to 80% of chronic conditions, yet the rich data generated by wearables, nutrition trackers, and electronic health records is still rarely used to support patients in their everyday decisions – between medical visits, in real life. 

With VitalHub, Liv Vital is already building an AI-augmented prevention platform that connects clinical diagnostics with lifestyle data and supports physicians with better insights for prevention, healthy aging, and longevity medicine. 

However, there is still a crucial missing piece:

Patients are left alone in their daily micro-decisions – “Is this meal ok for my cholesterol?”, “How do I adjust my movement while I’m sick?”, “What does my latest lab result mean for me right now?” – and existing chatbots are either too generic, not trustworthy enough, or not designed for sensitive medical contexts.

Recent advances in large language models and AI “copilots” open up the possibility of a Patient Copilot that acts as emotional safe and grounded intelligence: A conversational assistant that can translate the physician’s prevention plan into day-to-day, contextual guidance for the patient – available 24/7, but always safe, grounded in validated knowledge, and aligned with the treating physician.

To make such a system clinically usable, several open questions must be addressed:

1. How can we make the Patient Copilot safe and grounded? – Avoiding hallucinations, limiting scope, and reliably tying answers back to medical guidelines and the individual prevention plan.
2. How can we ensure that the AI is not biased? – Understanding and mitigating systematic differences in responses for different age groups, genders, or socio-economic backgrounds.
3. How can the Copilot build empathy and behavioural understanding over time? – Capturing interaction history, engagement patterns, and behavioural context so that the Copilot can adapt tone, level of detail, and motivational strategies for a more personalized interaction.

The TUM DI Lab offers an ideal environment to explore these questions at the intersection of AI safety, fairness, and human-centred conversational systems in a real-world healthcare setting.

Goal

Design and prototype a Patient Copilot for preventive and longevity care that:

• Interacts with patients in natural language and provides personalized, prevention-focused guidance,
• Is safe and grounded, by strictly basing its answers on curated medical knowledge and the physician’s prevention plan,
• Actively monitors and reduces bias, using a structured evaluation framework,
• Learns an empathy and behavioural profile of each patient over time to adapt its communication style and recommendations.The outcome should be a research-grade prototype, integrated with an anonymized version of the Liv Vital / VitalHub ecosystem, and a set of design & evaluation guidelines for safe, fair, and empathic AI copilots in preventive healthcare.

Requirements

Student Profile We welcome 4–5 motivated Master students with skills and interest in:

1. Technical skills 

• Python programming and experience with APIs,
• Basic knowledge of machine learning and/or natural language processing,
• Familiarity with LLM tooling (e.g. LangChain, OpenAI / other LLM APIs) is a plus,
• Optional: Frontend experience (React, Streamlit, or similar) for the chat UI.

2. Domain & methodological interests

• Strong curiosity for healthcare, prevention, and longevity medicine,
• Interest in AI safety, human-centred design, and fairness,
• Willingness to work with synthetic/anonymized health data under GDPR conscious workflows.

Opportunities This project offers a unique chance to:

• Work on a real-world AI product in preventive and longevity care, in close collaboration with a healthcare startup and clinical partners,
• Explore cutting-edge questions around safe and grounded AI, bias mitigation, and empathic conversational agents in a high-impact domain,
• Contribute to the next generation of Liv Vital’s platform – a Patient Copilot that could meaningfully help patients implement their prevention goals in daily life,
• Potentially co-author a scientific publication or white-paper and build a strong portfolio piece at the intersection of data science, AI, and digital health.

Apply to this project here