Data Berlin #47
World Cups, events, some new reading and more jobs for you
This World Cup had a bit of everything data visualizations (but we prefer the fan made), IoT (from the ball sensors), data quality checks (from VAR), and managers who wanted to have the results the way the wanted. At the end data don’t lie.
If you know, you know.
If you're not subscribed yet, it's free. Help us to reach 500+ readers, we gain nothing from this, but you can get newsletter like thisn in your inbox every week.
And if you're looking for your next role, our job board has 800+ open positions across 250+ companies at databerlin.net.
📖Interesting Readings
PostHog rebuilt its data warehouse on DuckDB, moving away from ClickHouse. The new architecture uses single-tenant DuckDB instances, DuckHog for local compute, DuckLake on S3, and Postgres Wire Protocol for tool compatibility – Why we rebuilt our data warehouse on DuckDB over ClickHouse
Hundreds of dashboards, most untouched, all running expensive queries every morning. The fix: 90 days with no views → archive (scream test), another 90 silent → delete. Keep a certified tier exempt, and celebrate deletions like launches. 9 out of 10 artifacts were dead – The 90/90 rule for the dashboard dumpster
A 7-step playbook for running evals on AI-on-data systems: how to measure whether an analytics agent, semantic layer, or schema change actually works on your warehouse. They want your email to give you this pdf guide – From Vibes to Evals: AI Analytics Eval Field Guide
The AI party isn't over, but the bill is coming. Semiconductors are in a bear market , hyperscalers are still spending like it's 2021, and the efficiency paradox means better models need less compute per output. Joe calls it the reconciliation phase: moving from possibility to accountability – The AI Party Isn't Over. But the Bill Will Show Up Soon
Five stories, one repricing: model capability is commoditizing, and the value has moved to everything around it. Training data (Grok exfiltrating repos), training environments (model-in-harness lock-in), distribution endpoints (Apple suing OpenAI over hardware), inference capacity (Anthropic's deadline extensions as market research), memory bandwidth (OpenAI's cache write surcharge). We always love Corey’s newsletter – The Model Stopped Being the Product
🚀 Shipping Now
dlt 1.29.0 — ClickHouse staging-optimized replace strategy with atomic
EXCHANGE TABLES, AWS Secret Manager config provider, BigQuery atomic replace viaWRITE_TRUNCATE_DATA, and DuckDB SQLAlchemy destination support.
Snowflake: CREATE OR ALTER now GA for 24 object types — Idempotent DDL for tables, views, procedures, functions, and more. No more drop-and-recreate for schema changes.
Snowflake: Clone dynamic Iceberg tables (GA) — Dynamic Iceberg tables can now be cloned, enabling zero-copy branching for Iceberg-based pipelines.
Databricks: Warehouse-level statement timeouts (Beta) — Set per-warehouse
statement_timeoutvia API, so different SQL warehouses can apply different query timeouts without cluster-wide settings.
BigQuery: Data governance tags (Preview) — Column-level security and data masking using Resource Manager tags. Also: Iceberg managed tables with partitioning and transactions are now GA, and project caps let you limit slots per reservation.
☕ Upcoming Meetups & Events
Jul 15 — PyData Berlin July Meetup (PyData Berlin) 🐍
Jul 18 — {Tech: Europe} x Almedia Hackathon | The Summer Lock-In ({Tech: Europe}) 💻
Jul 21 — AI Builders Berlin (AI Builders) 🏗️
Aug TBA — Data Berlin Summer Meetup (Data Berlin) 🐻
Aug 13 — Python in Berlin BBQ @ c-base (Python Users Berlin) 🐍
💼 This Week’s Job Picks
📊 Data Analyst
Compliance Expert Data Analyst – Qonto
Data Analyst - mid or senior level (F/M/X) – Ablefy
Business & Data Analyst - B2B2C – 1Global
SAP Business Intelligence Consultant – Arvato Systems
And more on databerlin.net/jobs/category/data-analyst
📈 Analytics Engineer
Senior Analytics Engineer, Marketing DSA – Vinted
Analytics Engineer (Decision Logic / Rules) _metrify – Enpal
Senior Analytics Engineer – Cosuno
(Senior) Analytics Engineer – Snowflake & Power BI – Europace AG
And more on databerlin.net/jobs/category/analytics-engineer
⚙️ Data Engineer
Data Engineer (m/f/d) – Berlin – Bees & Bears
Senior Data Engineer – GLS/NXT
Senior Data Platform Engineer (d/f/m) – Taxfix
Head of Data Engineering (m/w/d) – De Gruyter Brill
And more on databerlin.net/jobs/category/data-engineer
🤖 AI / ML / LLM + Data Scientist
Senior AI Engineer – Redcare Pharmacy
Deep Learning Engineer (m/f/d) - Audatic – Sonova
Machine Learning Engineer – Enfuce
Applied Scientist, Amazon Robotics R&D – Amazon / AWS
And more on databerlin.net/jobs/category/ai-ml
🎯 Leadership Roles
Senior Principal Engineer – AI Content Systems – Babbel
Director of Data & AI (Maternity cover) – Clue
Senior/Staff Backend Engineer - AI Compliance Tooling – Qonto
Head of Product - AI (F/M/X) – Ablefy
And more on databerlin.net/jobs/category/leadership
A lot of nice things end and so is this issue of our newsletter, and without a presindent trying to photobomb us. As you probably noticed this issues arrived quite late, apologies, we were getting a bit too much into the Summer vibe (can we still use vibe without a proper code review?).
IIf we were taking it easy on the newsletter front, our companion site kept going. This weekend we passed for the first time the 800 jobs on the board, and we found time to write abou the new RTO mandate in some companies in Berlin.
This is everything we have for this week. See you soon.
— Data Berlin team