heyPermit, Technical Co-founder

AI-assisted prototype designed to help environmental planning firms screen renewable-energy projects for biodiversity and infrastructure risks.

After six years at a startup builder, I wanted to try building my own company. The seed was planted at Antler , where I met my co-founder in "Berghain" - one of Antler's meeting rooms. We filled a whiteboard with subjects we cared about. One question clicked with both of us: could biodiversity conservation become a real business? We did not begin with software, but by talking to people. Our first interview was with a marine ecologist. Many more followed with scientists, conservationists, nature-tech founders, green-energy consultants, public servants, and people working in environmental planning. For almost a year, we followed every thread that caught our attention. We explored citizen science, marine ecosystems, biodiversity hotspots, soil health, eDNA labs, corporate sustainability teams, and climate reporting such as TNFD and CSRD . I learned far more about climate science and conservation than I had expected when we first stood in front of that whiteboard. But what stuck with us was the recurring pain-point of issuing permit applications for renewable-energy projects, saddled with lots of paperwork, and outdated processes. People reviewing an application had to collect data from project developers, find what is missing, and decide whether the project looked plausible. The slowest part was often the environmental impact assessment - or "Umweltverträglichkeitsprüfung," a German word almost as long as the process itself. We decided to focus on private environmental planning firms that prepared these assessments. We applied to several startup programmes in Berlin and joined KIEZ's Bridge-to-Market programme at TU Berlin's Centre for Entrepreneurship. From June to December 2024, we tested the problem, spoke with possible customers, and tried to find a business model. Six months later, we had to admit that we had not found one we trusted enough to continue. This was also the peak of the generative AI hype. SoTA vision models or VLMs were starting to beat traditional OCR tools at reading complex documents, while I was learning about GIS and digital maps. I built a prototype with Cohere 's open-source tools, OpenStreetMap , Leaflet.js, public biodiversity data from GBIF and iNaturalist , and public infrastructure APIs. The prototype gave planning firms a first look at whether a proposed development might affect a nearby protected area or biodiversity hotspot. It brought maps, project data, and environmental information together before a specialist began the full assessment. heyPermit did not become the company we had hoped to build, but the experience was deeply satisfying, and merits a longer blog post someday. I jumped into a field I barely knew anything about, spoke with people doing important work in this domain,  and came away much wiser than when I started.

https://web.archive.org/web/20241225011406/https://heypermit.com/