Ido Bronstein learned how to turn messy data into usable intelligence inside the Israeli military. Now he wants to do the same for companies trying to make artificial intelligence work.
Upriver, the startup he cofounded with Omri Lifschitz, has raised $14 million to help companies clean up and manage the business data that software tools need to work reliably.
Valley Capital and Hetz Ventures led the seed round, with participation from angel investors, including two of the founders of Cyera, the Israeli cybersecurity startup, last valued at $6 billion.
Three years ago, Bronstein was serving in an elite intelligence unit of the Israel Defense Forces, where he built software to turn raw data from multiple sources into information officers could use. He later managed a team of engineers that scaled that infrastructure.
Bronstein said the problem was similar to the one now facing data teams inside large companies. They also need to gather scattered information, make it usable, and deliver it to the people and systems that depend on it.
Companies are spending heavily to deploy AI, but those systems are only as useful as the data that feeds them. When that data is poor quality or spread across disconnected systems, Bronstein said, the tools may not perform as they should.
The typical fix has been to hire more people to maintain the underlying data, Bronstein said. Upriver is betting there’s a better way to scale: using autonomous agents to take on some of that work.
Most large companies have complicated data setups. A customer might pull data through tools like Fivetran, store it in Snowflake, orchestrate with Airflow, and visualize with tools like Tableau. Upriver’s agents plug into those systems and study how information moves through them. They can spot data that is incomplete, duplicated, or unreliable, and then help maintain the pipelines that move that data between applications.
Bronstein describes the software as a kind of superhuman data engineer. Rather than asking teams to manually check data quality, rebuild pipelines, or prepare datasets for tools, Upriver uses agents to perform some of that work on their behalf.
The company says its software is already used by Unity, DMGT, and Nimble, and that it has partnerships with Databricks and Snowflake.
The obvious question for Upriver is: why would Snowflake, Databricks, or another data giant not own this problem themselves? Bronstein argues that those platforms are part of the infrastructure Upriver is meant to work across, rather than replace.
Upriver may compete more directly with startups trying to automate the work of data teams with agents. Ardent AI, Definity, Databahn, and Matillion have all pitched versions of agentic data engineering, while Snowflake and Microsoft have already moved to buy startups in the space, acquiring TensorStax and Osmos, respectively.
Bronstein and Lifshitz met in Talpiot, an elite Israeli military program for technically gifted recruits who study advanced subjects while working on defense technology and research.
Talpiot has also become a well-known feeder into Israel’s startup sector. Its graduates include Check Point cofounder Marius Nacht; three of Wiz’s four founders, Assaf Rappaport, Ami Luttwak, and Yinon Costica; and founders of companies including Cyera, Classiq, and Via.
Upriver is betting that the same kind of training that produced some of Israel’s best-known startups can be applied to a more basic corporate problem: making sure companies can trust the data their tools depend on.
