Brief description
We are developing a shopping butler for supermarkets: an agentic AI assistant that supports users with their online shopping – from recipe and meal suggestions to product selection and filling the shopping cart. The butler is to be embedded in a supermarket’s app/website (e.g., REWE Online, EDEKA, dm) – not a separate chatbot, but part of the shopping experience.

This work focuses on the complete user interaction (UX/Interaction): voice-first communication, visual support, streaming/real-time feedback, and the transparent visualization of agentic actions.

Goal of the work

You will develop an interaction layer concept and implement it as a functioning prototype:

• The user speaks to the butler (voice-first).
• The butler responds primarily by voice, with optional visual additions.
• Agentic actions (e.g., product search, filtering, “Add to cart”) are visualized in the UI in a way that is understandable and trustworthy.
• The interaction feels fluid: streaming, short interim results, good turn-taking.

Tasks/Contents

• Design and implementation of an embedded UI concept for a butler in a supermarket app/website (present but not intrusive; shop remains visible)
• Development of UI patterns for agent action visualization (e.g., “shows/highlights” UI elements, status/timeline, confirmation/undo, security/transparency)
• Voice-first interface: microphone/listening UI, audio output, optional “silent mode” (captions/text cards)
• Streaming/real-time UX: integration of token/event-based streaming from the agent core (SSE/WebSockets), state rendering
• Improve turn-taking: Compare approaches (e.g., voice activity detection (VAD)/end-of-utterance/barge-in) and implement + evaluate one approach
• Integration into the existing system already being developed by other students

Technical topics (selection)

• Frontend engineering (e.g. React.js), state management, component design
• Real-time/streaming (SSE/WebSockets), event-driven UI
• Voice/audio: WebAudio, VAD/end-of-utterance, STT/TTS APIs
• Interface to the agent core: streaming, actions/events, UI rendering

What you should bring (must-haves)

• Extensive experience with frontend/UI projects (e.g. React or Angular) – ideally with real products/prototypes
• Strong software engineering fundamentals: Git, clean code structure, debugging, component architecture
• Experience with real-time UIs (WebSockets/SSE)
• High motivation, ownership, and desire for teamwork and integration into an existing system
• Very good written and spoken German and/or English skills (at least C level)

Nice-to-have

• Experience with voice/audio processing or speech interfaces (STT/TTS, VAD, streaming audio)
• Full-stack understanding (how FE ↔ BE ↔ DB interact)
• Initial experience with LLMs/agent systems (streaming responses, tool events)

What we offer

• A skilled, dynamic team with high standards and rapid iteration
• Product-oriented work with real user value and clear application
• Ample creative freedom in the interaction layer
• Opportunities for further work—depending on fit, also in a spin-off context

Application

Send us your profile (resume, transcript, GitHub/project links if applicable) and 3–5 sentences about:

• What have you already built in UI/frontend?
• What appeals to you about voice-first interaction & real-time UX?
• Mini challenge: Briefly outline your UI concept in 5–8 bullet points: How does the shop remain visible despite the butler, how does the butler display actions, and what does “silent mode” look like?

Contact Person

Andreas Morello, M.Sc.

Type of thesis

Master Thesis, Project Thesis, AI Project

Contact: