We are so back!!!
The autumn season kicked off at Svedea’s office in Stockholm, co-hosted with Snowflake. Summer was over, the community was back, and the room filled up with more people than usual, even welcoming a dear community member who had been part of the community from thousands of miles away and was finally experiencing his first DAIS event in person. Food and drinks were out from the start, and it did not take long for the catch-up conversations to get going.
Then the fire alarm went off. False alarm, as it turned out.
![IMG_7837.mov [video-to-gif output image]](https://substackcdn.com/image/fetch/$s_!_G9k!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9bb19b4a-c560-47e8-9f21-92b16c8bd5b3_800x450.gif "IMG_7837.mov [video-to-gif output image]")
Snowflake opened with a quick intro to their developer community and resources. From their forums and user groups to GitHub, Stack Overflow and YouTube, the message was straightforward: there are plenty of ways to get started and get help building on Snowflake.
Moving fast in an industry known for moving slowly
Rasmus Säfvenberg, Data Engineer and Product Owner at Svedea, opened with a line that set the tone for the whole evening.
Speedrunning Insurance.
Svedea has been building since 2010, growing from a single product to 13, with 600,000+ policies and 300+ employees across four offices in Sweden. It is a specialist insurer in one of the more regulated industries you can work in. And Rasmus was there to talk about what it actually takes to build a modern data platform inside that environment without cutting corners.
The speedrun framing was not just a gimmick. In a speedrun you move as fast as possible, but you follow the rules. As Rasmus put it, skips and cheating disqualify the run. For Svedea, working in a regulated industry, that balance between moving fast and staying inside the rules is not a choice, it is the job.
Why the platform had to change
Svedea needed a platform that scales. The one they had was not falling apart, but it was showing its limits.
Unclear lineage meant that tracing where a number came from took archaeology, as Rasmus put it. Ad hoc grants meant access had grown by exception over time, invisible until someone actually had to audit it. And compute was spread out, far from where the data lived.
How they did it
The first fix was treating onboarding as a config change rather than a project. One factory per source type, each system declaring itself, incremental by default. The idea being that a full reload should never happen by accident, it should be a deliberate choice.
The second was making access something you can actually see and manage. Access review became a pull request, 30+ declared roles and 60+ declared permissions, all version controlled. The deploy takes care of granting and revoking, so nothing grows invisibly in the background.
The third was about where compute lives. Before reaching for more of it, move it closer to the data. Every hop you remove is one less rule, one less endpoint, one less credential to justify to someone down the line.
What three months looks like
Three months in, the results were tangible. New sources that used to take three weeks to onboard now land in three days. Access is version controlled and reviewable as a diff. Daily deliveries running on schedule, inside the rules. Rasmus was honest about the downtime they had along the way, but the core of what they built held up.
Data in. Data out. Every day.
Take it home
Rasmus closed with three things worth carrying back.
Make onboarding a config change, because one factory per source type starts paying off fast once you are past a handful of systems. Declare access and let the deploy grant and revoke, so your access review is something you can actually read. And move the compute to the data before buying more of it, because every hop you remove is a rule, an endpoint and a credential you stop having to justify.
The best part of the first event back
Having Snowflake and Svedea on board brought more people through the door than usual, so Data & AI Stockholm moved straight from the talk into networking. No structured discussions this time, just people actually catching up after the summer. Conversations ran past the official end time and nobody seemed to be in a rush to leave.
Fun Fact: The evening had one unexpected moment. Rasmus had a Svedea Försäkringar campaign photo saved on his computer, and the face in it was none other than Vanessa, founder of Data & AI Stockholm, who used to work there. We loved it!
See You at the Data & AI Stockholm Summit
The Data & AI Stockholm Summit is on October 14th at Space Stockholm. To stay up to date with speakers, agenda and everything else, check the official Summit page.
Thank you to Svedea and Snowflake for hosting and making the evening possible, to Rasmus for a talk that gave the room plenty to think about, and to everyone who came out for the first event of the autumn. Good to have the community back in a room together.
Data & AI Stockholm 💙
Stay Connected
If you’d like to stay connected with the community and hear about future events, the best places to find us are: