Cloudera bought model-ops startup Verta to bolt AI onto its data platform

An operational AI platform that helps customers manage machine learning models, including the LLMs used in generative AI.

Almost $16 million raised in total (PitchBook), including a $10 million Series A in 2020.

Verta was founded in 2018 by CEO Manasi Vartak, who had created the open-source ModelDB database as a way to track versions of machine models while still in graduate school, later expanding the idea into Verta; co-founder CTO Conrado Miranda was once the machine learning lead at Twitter.

As companies moved toward LLMs, Verta evolved from a task-based model management platform into one geared toward managing large language models, acting as a control centre for the models. Cloudera — once the high-flying Hadoop startup that raised $1 billion, went public in 2018 and was taken private for $5.3 billion in 2021 — had launched a SaaS data lakehouse to chase cloud-native rivals like Snowflake and Databricks, which were adding AI organically and via acquisition.

On June 3, 2024, Cloudera announced it was acquiring Verta to strengthen its team and accelerate its operational AI capabilities, with CEO Charles Sansbury declaring 'the future of data management is AI'.

Cloudera's Hadoop-era business was losing ground to cloud-native data platforms that were already adding AI.

Verta had already pivoted its product to manage LLMs, giving Cloudera operational AI instead of a science project.

AI talent was scarce, and Verta's MIT CSAIL-rooted team came attached to the technology.

Verta's open-source ModelDB lineage gave the operational-AI approach a credible engineering foundation.

A model-ops startup survives pivots by riding each wave of model evolution — and exits to the incumbent that needs AI talent more than it needs time to build it.

As of the June 3, 2024 announcement, Verta's team and Operational AI platform were joining Cloudera to be integrated into its data platform; Verta ceased to operate as an independent startup.