Upriver secures $14 million to address the enterprise data issue in AI.

      The majority of enterprise AI projects do not fail due to poor models. Instead, their failure is often attributed to disorganized data, which includes broken pipelines, incompatible systems, and context that remains solely in one engineer’s understanding.

      Upriver, an Israeli startup, has secured $14 million in funding to automate the data cleanup process, believing that this essential yet mundane aspect is crucial for success in the AI era. The seed funding was led by Valley Capital Partners and Hetz Ventures.

      Notably, the angel investors include prominent figures from the data-tooling industry, such as Lew Cirne, founder of the observability company New Relic; Abe Gong, from the data-quality initiative Great Expectations; and the creators of the Israeli data-security startup Cyera.

      Upriver claims its platform is already utilized by Unity and the media organization DMGT, and has partnerships with Databricks and Snowflake.

      Describing itself as an "AI data engineering platform," Upriver functions as an agent that connects to a company's entire data ecosystem, including Snowflake, Databricks, BigQuery, Airflow, and dbt. It explores the data, constructs and verifies pipelines, repairs them when they malfunction, and transfers the "tribal knowledge" typically held by individuals into the system.

      The goal is for data engineers to no longer spend their time troubleshooting broken pipelines and integrating tools that are not designed to work together, but rather to focus on interpreting the data’s actual significance.

      The founders have a unique perspective, having spent ten years developing large-scale intelligence systems—work that Business Insider indicates was conducted for the Israeli military—before realizing that every organization using a modern cloud framework faces similar challenges.

      The new funding will be allocated to engineering, sales, and enterprise implementation. This timing aligns with a broader adjustment; after two years of investment in models and chips, companies are now evaluating the actual returns from AI, often finding that poor data is a significant issue. A wave of startups, including Capsa in private equity and Upriver within the data framework, is offering a similar core promise: that clean, reliable data is the critical factor determining the success of AI initiatives.

      Although the $14 million seed round is early stage and untested at scale, there is a growing consensus among investors that the foundational aspects are more important than the model itself.