About Sanvivo

Sanvivo GmbH is a Munich-based software & AI startup backed by Y Combinator (San Francisco) and Heal Capital (Berlin). We build pharmaceutical and medical software that helps pharmacies and doctors automate paperwork and operational processes — with a focus on the medical cannabis space. We are AI-native and already serve 100,000+ customers.

About the project

This is an Interdisciplinary Project (IDP) for a team of 2–3 students. You will design and ship production-oriented conversational AI agents that resolve real customer inquiries end to end, and extend their capabilities with system connectors (e.g., MCP servers) and well-defined operational guardrails.

Project goal

Deliver a production-oriented implementation of conversational AI agents that reliably handle a growing set of customer inquiries — combining our existing platform technology with new connectors and clear operational guardrails (evaluation, escalation, observability).

What you will build (work packages)

• Conversational agents: implement LLM-driven agents that respond directly to customer inquiries (support, product and process questions), including intent routing, tool-use, and human-escalation logic.
• Connectors & integrations: extend the assistant's reach by developing connectors — e.g., MCP servers — over internal APIs and data sources to grow the number of use cases handled.
• Guardrails & evaluation: define prompt and tool contracts, add automated evaluations, structured logging, and monitoring so agent quality is measurable and regressions are caught early.
• Delivery with stakeholders: work closely with the founding team, engineering, and the Head of Customer Support to iterate, validate, and ship increments.

Expected deliverables

• A working agent implementation (prototype or product increment, depending on the starting point).
• 1–2 new connectors/integrations or significant improvements (e.g., MCP-based services).
• Concise technical documentation — architecture, interfaces, operation and monitoring, known limitations — plus a final demo.

Who we are looking for

• Solid software engineering skills (backend and integrations; clean interfaces, tests, logging).
• Genuine interest in LLMs and conversational systems (prompting, tool-use, agent workflows).
• German at B2 level or better.

Logistics