Hi, I’m Sven Schultze, a PhD researcher at TU Darmstadt designing human-centered AI systems. #

I explore how people and AI agents can understand each other better: interfaces that explain what models see, multimodal assistants that act across devices, and infrastructure for the emerging agentic web where websites expose clear affordances to LLM-based agents.

What I’m working on #

• Agent-ready web design. VOIX lets developers describe what actions an agent can take on their site without brittle scraping. I test it with builders during multi-day hackathons and iterate with their prototypes.
• Explainable, multimodal interaction. From interactive aesthetics explainers to conversational robots in enterprise contexts, I build tools that keep people in the loop and make AI behavior legible.
• Pragmatic tooling. Open-source libraries like SymphonAI, tidy-env, and LiDAR localization experiments grow out of the day-to-day needs of researchers and practitioners I collaborate with.

Background #

Early on I was the on-call “family IT department,” which taught me how easily tech can alienate people. That experience still shapes my research: I look for ways to align AI systems with human expectations, expose what they can do, and keep people in control.

Recently I have:

• built explainable image-aesthetics tools so photographers can interrogate model judgments,
• collaborated with industry partners on deploying multimodal LLM assistants on humanoid robots,
• prototyped VOIX to let web developers declare agent affordances instead of forcing brittle screen scraping,
• created SymphonAI and tidy-env so other researchers can spin up multi-agent pipelines or interactive simulations quickly.

Outside the lab I design worlds as a Dungeons & Dragons dungeon master and produce music — both inform how I think about interaction, storytelling, and systems design.