
1
1
Germany
2+ years of experience
TUM Information Systems student building practical AI products and agentic workflows. Founder of Anytrace, an early-detection tool for VCs already used by Project A. Hackathon experience includes OpenAI Codex and TUM.ai, with a focus on fast MVP execution, product thinking, and polished demos.

SolarPinguis an autonomous AI sales swarm built to take the heavy lifting out of finding, evaluating, and closing solar leads. Instead of relying on a single rigid program, it coordinates a team of specialized AI agents that handle both sides of the sales funnel. For outbound prospecting, it automatically scans commercial areas via Google Places to find great candidates like logistics hubs or dealerships. For inbound traffic, it acts as a friendly digital intake assistant, welcoming customers through channels like Telegram and cleaning up messy website form submissions into structured profiles. The moment an address hits the system, the swarm goes to work on the technical details. One agent taps into the Google Solar API to instantly calculate the ideal panel count, estimated $\text{kWp}$ capacity, annual energy yield, and financial payback periods. At the same time, a vision agent uses multimodal open-source models via Featherless AI to inspect aerial satellite photos of the roof, checking for tricky obstructions like chimneys or vents to make sure the solar potential is actually real. If a property checks out and hits the quality thresholds, SolarPingu immediately puts together a customized, professional PDF proposal. To wrap it all up, the system can even jump on the phone autonomously using Twilio and Speechmatics, running natural conversational workflows, handling objections in multiple languages, and streaming live transcripts straight to a central dashboard. With built-in fallback options and smart mock modes, the entire setup is incredibly resilient, keeping the pipeline moving smoothly without breaking a sweat.
19 May 2026