Coding for days, hardly any sleep and a first prototype: This is how Amerigo Velletti’s first week at the summer camp “Founder scene is looking for super founders” went. His idea: to build an “idealo for second hand”.
Sleep? Rather in short supply. Motivation? 100 percent. This is how Amerigo Velletti describes his first days at our startup camp “Start-up scene is looking for super founders”. Eight selected fellows have ten weeks to develop consumer apps using artificial intelligence.
Amerigo has thus fully started the life of a founder. The 21-year-old moved from Munich to Berlin especially for the camp – and currently spends almost every waking hour on his idea. “I still haven’t met my roommates because I always don’t come home until 1 a.m. at the earliest. And I always leave before they’re even awake. So, I always see their shoes in the morning, but I have no idea who they actually are.” He says he didn’t take a break on the weekend either. Instead, he continued working on his idea.
“Startup scene is looking for super founders” is a ten-week startup fellowship from Gründerszene. Eight selected fellows develop consumer startups in Berlin using artificial intelligence – from the idea to the first product. During the program, the fellows work together in the Axel Springer high-rise and are accompanied by experienced entrepreneurs, investors and experts. Partners like OpenAI, Vercel, Dash0 and DHL support them with technology, know-how and mentoring. Gründerszene documents the entire journey with articles, videos and social media content – and shows up close how the next generation of startups is being created today.
Ideal for second-hand products
His idea: an AI-supported search engine for second-hand products – cross-platform. “Quasi idealo for second-hand. That’s my slogan, which I also used when applying,” he says in an interview with Gründerszene. His app is intended to bundle offers from platforms such as Classifieds, Vinted, Back Market or Refurbed. Users simply specify what they are looking for, the AI searches different marketplaces and delivers the most suitable hits.
Amerigo is not just concerned with the cheapest price. The condition also plays a crucial role, especially with used products. “When I buy anything second-hand, I always look at the condition.” At the start he is therefore concentrating on electronic devices.
His target group are people who want to shop sustainably. In the long term, he wants to simplify the entire purchasing process: “It would actually be my dream to build an end-to-end funnel so that the user has to spend less time shopping, so to speak.”
This is how Amerigo wants to optimize the search functions
The idea came from an everyday problem. “For example, if I search for a specific lamp on the eBay classifieds app, I am offered 18 different lamps before I see the lamp that I am actually looking for.” This is exactly where AI is supposed to help. “I think a categorizer like this, especially an image categorizer with embedding models like the ones you can build now, is a very current fit that can also be implemented very easily from a technical point of view.”
Read too
The first steps to the comparison platform
So far, Amerigo has hardly needed any technical support. “I’ve been programming for a few years.” He brings a strong technical background. In addition to his work as a founding member of Vakanzio, where he worked on AI agents, agent-based RAG pipelines and full-stack automation with Python, LangChain and React, he completed his bachelor’s degree in business informatics at the Technical University of Munich with a grade of 1.0.
He therefore spent the first days of the summer camp primarily building his product. “I’ve been coding for three days. So really not that much has happened yet.”
His most important tool is Codex (note: OpenAI is a partner in our fellowship). Additionally, he uses Superset as an orchestration framework to allow multiple AI agents to run in parallel without interfering with each other. For the front end, he uses, among other things, Google Stitch, which generates user interfaces and production-ready front end code from text prompts, speech or sketches.
In addition to the technical implementation, he obtained feedback early on. He talked to friends about his idea and contacted people who were working on similar products. At the same time, he analyzed existing providers such as eBay, Rebuy and other second-hand platforms.
Now it’s all about testing. Amerigo has already taken action on this too: it has contacted the platforms whose offers its users would later be redirected to.
Read too
The biggest challenge: getting the right data sets
This is exactly where he runs into a problem. “Actually, what I want most is to have their data. So to get started, the best thing I need is a high-quality data extract so that I can start testing with real examples.”
In the long term, his app should be able to do much more than combine offers from different platforms and redirect users. His vision is so-called Autonomous Buying Recommendations: AI agents that not only recommend suitable products, but also independently accompany the purchasing process until shortly before completion.
But that’s exactly why he needs high-quality data sets – and they’re hard to come by. “So far there has been little return,” he says. He received a small data extract from Refurbed. He was also able to organize data sets via eBay. The exchange with Vinted is more difficult: “Vinted answered me in French, even though I asked English questions. I’m a bit stuck there.”
Nevertheless, he continues to work on connecting additional platforms to his application. There is already a first prototype. It currently runs under the name Remtio – even if Amerigo is not happy with the branding yet. He’s harsh on the current state of development: “Well, I’m not really happy with where things are right now, but maybe I just have very high standards.”
The dependence on external partners in particular costs him time. “If you know how fast things can actually happen, but you get attached to things like that, then it gets frustrating.” He also reveals that he is in contact with Idealo from the Springer network. What a possible collaboration there could look like will become clear in week two of the camp. On top of that, he wants to further improve the prototype. His goal for week two: His AI should accompany the purchasing process a decisive step further. Until now, the agent could recommend suitable products and put them in a shopping cart. In the future, however, it should guide users seamlessly through to the actual purchase.
Read too
“Stick to the basics.”
The most memorable conversation of the first week was with Dash0 co-founder and CEO Mirko Novakovic. “That was really incredibly impressive.”
Above all, he takes one insight away from the exchange: Success does not come from one brilliant idea, but from consistent work on the product. “The key learning I took away from the week was definitely to stick to the basics and just keep going.” For him that means starting on time every morning and coding for hours. “Work on the product until it meets your own quality standards. And then be a little proud of yourself.”
“Start-up scene is looking for super founders” powered by Dash0, DHL, OpenAI & Vercel.