Data Berlin #41
The semantic layer won. Also, Berlin companies, please post salaries.
The EU Pay Transparency Directive came into force in Germany on June 7. Already the data is telling a story: only 7% of roles on the board include usable salary ranges. We’re tracking how Berlin companies respond, and what it signals when they don’t. Read our take.
On the tooling side, last week Anthropic published how they run self-service analytics with Claude internally, and Snowflake Summit was three days of CoWork, Managed Agents, Cortex. The pitch: ask the platform, not the data engineer. Whether that changes how Berlin teams hire is still an open question, one we’re exploring live on June 18 (which is unfortunately full booked).
What’s new on our job board
We are now at 700+ roles on databerlin.net/jobs. The improvements made at the end of the last month really changed the curve of our spending (the goal is to stay below the free credit of 12$ dollars per month. With an average spend of ~0.16 in June versus the ~0.42 of May.
Follow us for other cheap tricks to run your data platform 😉.
📖Interesting Readings
Snowflake’s Summit features only make sense as a connected architecture, which matters because understanding how Horizon Context and Cortex Sense feed the activation layer (CoWork, CoCo) clarifies what’s genuinely new versus rebranding. – Snowflake Summit AI stack, mapped
A governed semantic layer (mentioned 14 times as the mandatory foundation) paired with pre-built skills pushed accuracy from 21% to 95%, which matters because metric governance turns out to beat prompt engineering as the primary lever for reliable AI querying. – How Anthropic enables self-service data analytics with Claude
Format and source type produce a 30x throughput spread on identical hardware (Parquet at 170 GB/hr vs JSON at 4.6 GB/hr), which matters because tuning your ingestion engine is the wrong lever when the real bottleneck is schema inference or API rate limits. – How fast is dlt? A benchmark
Most AI products sit in a zone where directional accuracy is acceptable, which matters because deploying a model calibrated for “good enough” into a context that demands exact precision is a category error that no amount of fine-tuning fixes. – Cooperating with failure
The field adopted CI/CD and source control from software engineering but skipped the shared vocabulary for stateful problems like retries and idempotency, which matters because without named abstractions every team builds equivalent solutions without knowing it. – We borrowed the tools, not the patterns
🚀 Shipping Now
🚀 Shipping Now
PostgreSQL 19 Beta 1 is out, opening a public testing window ahead of the stable release this autumn.
ClickHouse rolls out 26× faster join performance on analytical workloads with a rewritten join execution engine.
MotherDuck adds SQL access to Obsidian vaults, letting agents query note-based knowledge bases directly via DuckDB.
Bruin ships ingestr v1, a Go rewrite of its data ingestion CLI claiming best-in-class throughput.
Omni launches AI Hub, a command center for tracking AI usage in BI and catching semantic model quality issues before they reach production.
dbt Core v2 is here, built on the same Rust foundation as Fusion and relicensed to Apache 2.0, with faster parse times and a unified runtime.
There are many posts and articles out there about the Snowflake Summit and all the things they announced. We leave that as homework for the interested reader.
☕ Upcoming Meetups & Events
Jun 10 — BLISS · BLISS x QuantCo Workshop 📐
Jun 10 — PostgreSQL Berlin · PostgreSQL June Meetup 🐘
Jun 16 — Bob DACH Developer Community · Session & Coding Watch Party💻
Jun 16 — Berlin Power BI User Group · Introduction to DAX User Defined Functions (UDFs) 📊
Jun 18 — Data Berlin · Agentic Analytics Meetup 🐻
Jun 19 — AI-Memory Hackathon 🚀
Jul 8–10 — WeAreDevelopers · World Congress — CityCube, Messe Berlin; 20% discount with Community_DataBerlin 🎪
💼 This Week’s Job Picks
📊 Data Analyst
Mid Data Analyst – Finance – SumUp → Apply
Senior Data Analyst – Yepoda → Apply
Data Analyst HR 💰 – Redcare Pharmacy → Apply
And more here.
📈 Analytics Engineer
Senior Analytics Engineer – Run & Grow – SumUp → Apply
Analytics Engineer, Finance – Vinted → Apply
Senior Analytics Engineer – Finance & Operations – PERGOLUX → Apply
Senior Analytics Engineer – Distribusion → Apply
And more here.
⚙️ Data Engineer
Data Platform Engineer (Mid-Level) – FlixMobility → Apply
Senior Platform Engineer – Kafka – emnify → Apply
Site Reliability Engineer – Data Platform – N26 → Apply
Senior Data Engineer (Modern Data Platform & AI) – AroundHome → Apply
And more here.
🤖 AI / ML + 🧪 Data Scientist
Senior DS/ML Engineer – Financial Crime – SumUp → Apply
Senior ML Ops Engineer 💰 – Redcare Pharmacy → Apply
AI Engineer – Raisin → Apply
Senior Data Scientist – Just Eat Takeaway → Apply
And more AI/ML roles here and Data Scientist roles here.
👔 Leadership
Lead Credit Risk Data Scientist – Billie → Apply
Manager, Payments Risk & Analytics – PayPal → Apply
AI Tech Lead – JTL-Software → Apply
Team Lead Data Engineering – idealo → Apply
And more here.
Two things landed at once this week: a law that should change how Berlin companies post jobs, and a wave of conference announcements about AI doing the analytics work itself. Whether either sticks is the question. Hit reply, we’d like to know which AI tools your team is actually running, and whether your employer posted a salary range yet.
— Data Berlin team