Start-Up

PARAT builds the offline cookie for physical commerce. Online, every shopper is identified by a cookie; offline, a store or brand sees nothing about who came, who came back and who stopped coming. We put the customer's identity into a pass in Apple or Google Wallet, so there is no app to download, and every visit, stamp and scanned receipt becomes a structured event. That gives stores and brands first-party customer data where there was none before.
On top of that data, an agent, built specifically for the wallet channel, recognizes the individual customer, picks the right moment, and sends a personalized offer through the wallet, WhatsApp, or iMessage, with the business approving before anything goes out. Our self-serve product Bonice is live in about 120 paid locations across Germany, Austria and Switzerland, with more than 14,800 wallet passes issued, and our customers range from independent cafés to bakery and franchise chains with 30 to 60 stores.
We are a team of seven, founded in 2025 and funded by the Berliner Startup Stipendium and AI Nation (TU Berlin), and we are currently raising our seed round and negotiating our first enterprise rollout with a large German grocer. We build AI-native: the product was built by a very small team using AI throughout the stack, and much of our go-to-market runs on AI agents as well. An intern joins a small team where their work goes into production and in front of real customers within weeks.

The intern works with the CPO on the agentic activation layer of our product. We hold visit, stamp and receipt data from tens of thousands of end customers across bakery, gastronomy and retail chains, and the open question is how an agent should use it: which moments are worth acting on, which message fits which customer, and how to prove the effect on visit frequency and revenue.

Concrete tasks: analyzing customer behavior across our live chains (visit patterns, drop-off, redemption rates, weekday and weather effects) and turning findings into reports our customers actually read; designing and evaluating activation rules and agent prompts, measuring their effect in production; prototyping product features with AI tools such as Lovable and Cursor together with the founding engineer; joining customer calls to understand how a bakery chain or a franchise HQ reads the data; and contributing to the product roadmap in Linear. The intern ends the summer with at least one analysis and one feature that are in use by real customers.

Try Bonice as a customer (bonice.de) and think about what data a store gains from it. Basic familiarity with SQL and one scripting language.
Optional reading: how Apple Wallet and Google Wallet passes work, and one or two pieces on agentic AI in CRM. We send a short reading list and data access before the start.