Upriver secures $14 million to address the data challenges faced by AI in enterprises.

      Most enterprise AI initiatives do not fail due to poor model performance. Instead, they falter because the data that supports them is chaotic, characterized by broken pipelines, incompatible systems, and contextual knowledge that is often confined to a single engineer's understanding.

      Upriver, an Israeli startup, has secured $14 million in funding to automate data cleanup, believing that this crucial yet unexciting layer is where the real competition in the AI landscape is determined. The seed funding round was led by Valley Capital Partners and Hetz Ventures.

      The list of angel investors is particularly noteworthy, featuring prominent names from the data-tooling sector: Lew Cirne, the founder of the observability giant New Relic; Abe Gong from the data-quality initiative Great Expectations; and the founders of the Israeli data-security unicorn Cyera.

      Upriver claims to be utilized by companies like Unity and the media group DMGT and partners with Databricks and Snowflake.

      Positioning itself as an "AI data engineering platform," Upriver acts as an intermediary that connects to a company's entire data stack—including Snowflake, Databricks, BigQuery, Airflow, and dbt—explores the data, builds and validates pipelines, repairs them when they malfunction, and captures the “tribal knowledge” that typically resides in people's minds.

      This approach promises to free data engineers from the tedious tasks of troubleshooting broken pipelines and integrating disparate tools, allowing them to focus on deriving meaning from the data instead.

      The founders have a unique background; CEO Ido Bronstein and CTO Omri Lifshitz dedicated a decade to building large-scale intelligence systems, work that Business Insider indicates was conducted for the Israeli military, before realizing that many companies using modern cloud stacks face similar challenges.

      The new funding will be allocated towards engineering, sales, and enterprise deployments.

      This comes at a time of a broader reevaluation, as companies reassess their investments in models and chips after two years of spending. A common discovery is that AI falters due to poor data quality. A wave of startups, from Capsa in private equity to Upriver within the data stack, is emphasizing the same fundamental message: clean and reliable data is critical for transforming an AI pilot into a successful application.

      While this $14 million seed round is still early and untested at scale, the belief that foundational elements are more significant than the model itself is gaining traction among significant investors.