Exclusive: Voltron Data acquires Claypot to unlock real-time AI

Today, San Francisco-based Voltron Data, a startup providing enterprises with a modular and composable approach to building systems for data analytics, confirmed to VentureBeat that is acquiring real-time AI platform Claypot. The terms of the deal were not disclosed.

The acquisition will bring Claypot’s entire team under Voltron’s umbrella, enabling the latter to expand its offering to include real-time analytics. This will give enterprises a way to unlock value from the freshest data and ultimately help with use cases like fraud detection and personalization.

“Our two teams have been working together for the last eight months to build the first streaming data backend… and we are thrilled to bring all the Claypot AI founding members to Voltron Data to continue their work in streaming and unifying real-time and batch data processing. We’re bullish about what we’ll be able to achieve together,” Josh Patterson, the co-founder and CEO of Voltron Data, told VentureBeat.

Voltron Data brings the best of open source together

Founded in 2022 and backed by $110 million in funding, Voltron Data seeks inspiration from the fictional animated series Voltron, in which small piloted robots combined into a larger unified mecha.

It brings multiple open source technologies, including Apache Arrow, Apache Parquet and Ibis, together to help improve data access and analytics.

The idea was simple: give teams a way to select the best component at each layer of the data stack and build modular systems that could be augmented to meet evolving organization needs.

Last month, the company debuted its Theseus distributed query engine to accelerate CPU-dependent data preprocessing tasks – ETL, feature engineering and transformation – with a cluster of Nvidia GPUs and other hardware accelerators. The move gave a way to do both data preprocessing and AI/ML on GPUs, unifying data analytics and AI pipelines on the same infrastructure. Now, building on this work, it is acquiring Claypot AI.

Based in San Francisco, and run by former Nvidia and Netflix engineers Chip Huyen and Zhenzhong Xu, Claypot allows enterprises to leverage the freshest data to make more accurate predictions, get faster insights into the production environment and speed up model iteration to adapt to data distribution shifts.

However, it doesn’t just work with streaming data. The offering combines both streaming and batch data processing to handle different enterprise needs — optimizing for latency, cost and correctness at the same time.

“Claypot works with both streaming and batch data, allowing you to choose the best-performing data for each use case. If a one-hour delay is good enough, we’ll switch to batch. Things are changing fast and you need a reaction in milliseconds? We’ll switch to streaming,” the company notes on its website.

With the technology and team of Claypot coming on board, Voltron’s modular systems, which previously focused largely on local and batch data, will be able to deliver analytics from streaming data, using the same open standards. This means enterprises will get real-time AI, feature engineering and MLOps capabilities powered by Theseus as well as open-source products.

“We will give system builders more choices as we bring on the Claypot AI team. We will work together to extend our vision of composability and hardware acceleration further into real-time data and machine learning operations (MLOps),” Patterson said.

Goal to break down technology silos

While it remains unclear when exactly the integration of Claypot will be executed, the company is bullish on the prospect of breaking down technology silos across data analytics and AI and has more efforts in the pipeline.

“We have more foundational work to do…Two of the highest-value use cases we see today are preprocessing structured and semi-structured data for ML/AI models and operating on real-time data streams with low latency requirements. Thesues is designed for workloads too large and time-sensitive for Spark and Presto. We don’t have any (other) acquisitions planned at this time, (but) we are always looking for top talent aligned with our vision to help accelerate our growth,” the co-founder added.

Since its launch, Voltron Data has roped in “numerous” customers across major industries including the federal government, where it is partnering with two agencies to accelerate query performance on petabyte-sized data. In November 2023, the company also partnered with HPE to integrate Theseus in HPE Ezmeral unified analytics software. 

“Our customers are experiencing increasing benefits of our accelerated solutions, expanding to larger, more critical data problems with ease,” Patterson said.