mindit.io

Two customer records scored 0.9998. They were two different people. Same family email, same address, different phone numbers. The merge was declined and the pair went to a reviewer. That’s the problem we solved for a luxury brand selling worldwide. Its clients sat across three systems (online shop, store tills, CRM) with nothing connecting records that belonged to the same person. We built the customer matching on Databricks, with one rule from the brand: join two records only when the evidence is conclusive. That meant treating shared emails and phones with suspicion, and using a stricter threshold to merge identities than to group them. The result: about 84% of records match automatically, 14% stand as unique clients, and around 2% go to a reviewer. Cristian Badet explains how it works in the full article. Link in the comments.

FINMA and BaFin have not written an AI rulebook for banks. They read AI into the rules banks already follow: model risk, outsourcing, ICT risk under DORA. So most banks already own the controls. What they lack is proof, per AI system, that the controls ran. The ECB’s findings from 13 supervised banks: → About half had a dedicated AI policy → None had automated validation tools for AI models → None had a fallback for when an AI system fails Our test: can a risk officer answer the auditor’s question with a query instead of a slide? Full article in the comments.

Full house at Databricks Bucharest User Group #3: 110+ people at The Marmorosch Bucharest | Autograph Collection, and we're still thinking about the conversations from the room. Thank you to our speakers for sharing real work with the community: 💳Mircea Idriceanu (BCR) and Alexandru Puiu on migrating the biggest EDW in CEE to Databricks 🧱Vlad-Alexandru Mihalcea and Burak Seyid (Databricks) on governance, from cost optimization to AI transparency ⛽Michael Reisinger and Andrei-Cosmin Ciocan (OMV Petrom) on AI-powered energy trading And thank you to everyone who came, asked questions, and stayed to talk afterwards. A user group is only as good as the people in the room, and this one was a good room. As co-organizers of the largest Databricks meetup in #Bucharest, we're already working on what comes next. Stay tuned for our upcoming events. Want the full story from the evening? Article link in the comments.

Databricks Bucharest User Group #3, sponsored by mindit.io, is happening on September 30, and two of our leading engineers are taking the stage. Alexandru Puiu (CTO) joins Alexandru Antonescu (Executive Director, Data, Analytics & AI at BCR) to break down migrating the biggest EDW in CEE to Databricks, and what modernization actually takes at that scale. Vlad-Alexandru Mihalcea (Lead Data Engineer) teams up with Burak Seyid (Solutions Architect at Databricks) on governance: cost optimization and AI transparency, and how to keep a Databricks deployment under control as it scales. Also on the agenda: Michael Reisinger, Digital Trading Manager at OMV Petrom, and Andrei-Cosmin Ciocan, Data Engineering & AI Technical lead at OMV Petrom, will be covering Al-powered energy trading.on AI-powered energy trading. 📌The Marmorosch Bucharest | Autograph Collection 🗓️September 30, 2026 | 5:45 PM – 9:00 PM (GMT+3) Register here to join our event: https://lnkd.in/dBc2-6sW

Did we mention how much we enjoy talking to our colleagues and customers about data? Most teams underestimate how complex a data migration really is, often by 2 to 3x. We built a quick assessment to help you see where you actually stand before it becomes a costly surprise. 15 questions. Less than 4 minutes. One readiness score. Take it here: https://lnkd.in/enAZpw9w

Rule-based recommendation engines are costing DACH retailers sales. AI-powered personalization delivers 25-35% higher conversion, and real-time recommendations beat segment targeting by 40-60% on high-frequency categories. The gap versus Amazon and Zalando is measurable, and closing it doesn't require a full data overhaul. A focused rollout on email and on-site recommendations, using data most retailers already have, can show ROI in 4-6 months. Our new guide covers the four components of a production-ready personalization stack, which model to pick based on your data maturity, and how to build in DSGVO-compliant consent from day one. Read it here: https://lnkd.in/dFZ5UTwp mindit.io builds end-to-end personalization platforms for retail, banking, and insurance across DACH, UK, and BENELUX.