On a late winter morning in Berlin, Serhii Sokolenko and Brad Heller did what founders do when they've just closed a funding round: they announced it to the world. The March 13, 2026 press release was careful, measured—$6.4 million across pre-seed and seed rounds, investors with recognizable names, the usual talk of "next-generation infrastructure."
What makes Tower.dev's pitch unusual isn't the money. It's the wager embedded in it.
The two former Snowflake engineers are building on what they see as a quiet linguistic shift in the data world—away from SQL-first tools like Airflow and traditional ETL pipelines, toward Python-native ecosystems built around Polars, dbt, dlt, and LangChain. Tower wants to be the infrastructure layer that makes that transition seamless. Or, as the founders frame it with characteristic startup bravado, "the Databricks of the Python Era."
Whether the market agrees is another question entirely.
The Money and the Believers
Speedinvest led the seed portion of the round, with DIG Ventures anchoring the pre-seed. The syndicate expands from there: Flyer One Ventures, Roosh Ventures, Celero Ventures, Angel Invest. Standard fare for an early-stage European infrastructure play.
More telling, perhaps, is the angel roster. Jordan Tigani, who runs MotherDuck. Olivier Pomel of Datadog. Ben Liebald, lately of Harvey.ai. Maik Taro Wehmeyer from Taktile. These aren't passive check-writers—they're operators who've built or scaled data platforms themselves, the kind of backers who signal that something technical is worth paying attention to.
Still, conviction from insiders doesn't guarantee product-market fit. It just means the problem is real enough to warrant a closer look.
What Tower Actually Does
The technical pitch is dense but coherent. Tower merges three layers—orchestration, compute execution, and lakehouse storage—into what it calls a "multi-tenant control plane." The infrastructure handles both serverless and self-hosted workloads, leans on Apache Iceberg for lakehouse management, and integrates with incumbents like Snowflake and Spark rather than trying to replace them outright.
Recent feature rollouts include automatic retries with backoff policies (essential for flaky data pipelines), webhooks that trigger on run events, and flow graphs for mapping complex workflows—including those generated by AI agents, a nod to where the founders think the puck is headed.
The underlying thesis: as teams generate more data transformations and orchestration logic in Python—sometimes written by LLMs rather than humans—they need infrastructure that treats Python as a first-class citizen, not a bolt-on scripting layer.
It's a reasonable bet. But it requires data engineers to rethink workflows they've been running for years.
The Pedigree
Sokolenko and Heller's resumes read like a tour of modern data infrastructure's greatest hits. They met at Snowflake, where both worked on control plane performance—unglamorous but critical work. Before that? AWS Databases, Google Cloud Dataflow, Databricks, Puppet. Heller, the CTO, has founder scars already: a previous exit to Puppet, among other ventures.
They started Tower in 2024 with two design partners: dltHub and Taktile. Both have since migrated production workloads onto the platform—a signal that whatever Tower built, it at least solves problems for early adopters. A case study from dltHub describes it as a "portable Python runtime," which in plain terms means it bridges the gap between tinkering on a laptop and running code at scale.
That's the kind of pain point that resonates if you've ever wrestled with containerization, environment inconsistencies, or the "works on my machine" problem.
What Comes Next
The team stands at twelve people. They're hiring—principal and senior engineers in Berlin and London, go-to-market roles in London and the U.S. The expansion feels measured, not frantic.
Tower is also investing in community visibility. They're sponsoring Iceberg Summit in San Francisco later this year and exhibiting at Big Data & AI World in London. Smart moves for an early-stage company trying to carve out mindshare in a crowded space.
The feature roadmap leans into enterprise concerns: self-hosted runners for sensitive on-premises data, unified observability, webhook integrations that let Tower play nicely with existing orchestration tools. These aren't flashy product announcements. They're table stakes for selling into regulated industries or security-conscious enterprises.
The Python Question
The real test isn't whether Tower can build good software—early traction suggests they can. It's whether the market is actually ready to reorient around Python the way Sokolenko and Heller believe it is.
SQL isn't going anywhere. Airflow, for all its quirks, powers thousands of production pipelines. Databricks and Snowflake have enormous installed bases and aren't standing still. Tower's bet requires persuading teams that the friction of migration is worth the promise of Python-native simplicity.
The investor lineup suggests some influential people think that bet is sound. But infrastructure shifts happen slowly, even when they're inevitable. And in a market already thick with orchestration platforms, observability tools, and lakehouse vendors, Tower will need more than a compelling technical story. It'll need distribution, customer success stories, and the kind of grassroots adoption that turns a clever product into an industry standard.
For now, the founders have bought themselves runway. What they do with it—and whether the Python wave they're surfing proves real—will determine whether Tower becomes the Databricks comparisons it's courting, or another well-funded also-ran.
Either way, it's a story worth watching. Infrastructure bets this ambitious tend to clarify themselves quickly.
