Inside SereneDB: How this Berlin startup built the world’s fastest database
Tech Funding News Sofia Chesnokova
Berlin startup SereneDB says its database finds records in about 200 microseconds. It’s claiming the fastest search in the world, backed by open benchmarks.
Based on reporting by Tech Funding News, Sofia Chesnokova — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
SereneDB, a Berlin startup founded in 2025, is making a very specific sort of brag: it says its database can find a single record among billions in about 200 microseconds. Alexander Malandin, the company’s co-founder and CEO, says that put SereneDB ahead of Elasticsearch and ClickHouse in benchmark tests on a 10-billion-record dataset. Against ClickHouse, he says, the gap was about 10x.
The company is not asking anyone to take its word for it. It built an open benchmark called SearchBench on the same model as ClickHouse’s ClickBench, and Malandin says ClickHouse’s own team was the first to respond, submitting changes that SereneDB folded into a retest. That matters because this is one of those claims that lives or dies by the benchmark page, not the pitch deck. SereneDB now says, bluntly, that it is the fastest search in the world.
The product itself is open source and Postgres-compatible, released under the Apache 2.0 license. SereneDB was founded by Malandin, Andrey Abramov, and Valery Mironov, and its engine mixes vector search, full-text search, and analytics so companies do not have to bolt together separate indexes, caches, and dashboards. Malandin’s analogy is simple: most database indexes work like a table of contents, while SereneDB tries to behave more like a glossary, doing more work when data is written so queries can return almost immediately.
That speed pitch is aimed less at giant, already-entrenched enterprises and more at what Malandin calls vertical AI: small teams using AI agents across very large datasets. His argument is that these teams can start looking like enterprises in their data needs, even if they do not look like them on a payroll sheet. He points to examples like a SereneDB developer who built a game analytics platform from 30 million public Valve match files in two weeks, and says similar patterns should show up in biotech and genomic research.
The company itself is tiny by comparison. It has eight people, most of them engineers, all in Berlin. In December 2025, it raised $2.1 million in pre-seed funding from Entourage and High-Tech Gründerfonds, and its early commercial traction has come through a small number of resale partners found by word of mouth. SereneDB is planning another funding round in early 2027, with two products in mind: managed hosting and an on-premises licence for customers who cannot use the cloud for compliance reasons.
My take — AI-written commentary, not fact-checked reporting
This is the kind of startup story that actually deserves attention: not another AI wrapper, but a company attacking the plumbing. Berlin keeps producing serious infra teams that prefer code to hype, which is refreshing. The open-source route helps too; if the benchmark is real, the market can argue with it in public instead of in a sales deck.
Read more about this at: Tech Funding News
Related stories
Keenable AI Open-Sources NEEDLE: A Live Search Benchmark That Rebuilds Its Query Set Every Hour
MarkTechPost · 4 weeks ago ·
2
Perplexity’s AI agents helped build a database. They weren’t allowed to run it.
The New Stack · 1 week ago ·
47