Eventual Emerges From Lyft’s AI Program to Power the Future of Multimodal Data Processing

TL;DR

• Eventual, co-founded by two former Lyft engineers, is building Daft, a Python-native multimodal data processing engine.
  
• Daft processes unstructured data types—text, video, audio, images—in a unified, high-speed framework.
  
• Originally built for autonomous vehicle data, Daft now supports clients like Amazon, CloudKitchens, and Together AI.
  
• Eventual recently raised $20M Series A led by Felicis, after a $7.5M seed from CRV.
  
• Backers include Microsoft’s M12 and Citi Ventures, with the funding aimed at expanding the open source Daft and launching enterprise tooling in Q3 2025.
  
• The rise of generative AI has created surging demand for multimodal infrastructure, with projected 35% CAGR through 2028.
  

Origin: From Autonomous Vehicles to Open Source Innovation

The foundation for Eventual was laid inside the engineering labs of Lyft’s autonomous vehicle division, where co-founders Sammy Sidhu and Jay Chia faced an overwhelming problem: autonomous cars generated vast volumes of unstructured, multimodal data—from 3D LiDAR scans to video feeds, audio snippets, and metadata.

Yet there was no unified platform capable of ingesting and processing this data efficiently.

“We had PhDs spending 80% of their time on infrastructure instead of applications,” said CEO Sammy Sidhu in a recent interview with TechCrunch.

In response, Sidhu and Chia developed an internal tool to process this rich and varied data. Its success later evolved into the foundation for Eventual, a startup founded in 2022—months before ChatGPT introduced the world to the generative AI boom.

Daft: SQL for the Unstructured Data Era

At the heart of Eventual’s platform is Daft, a Python-native open source data engine designed to process multimodal inputs—audio, video, text, images—all under a unified architecture.

Sidhu likens the engine’s ambition to that of SQL’s historic impact on structured data.

“Our vision is for Daft to be to unstructured data what SQL was to tabular data,” said Sidhu.

The open source version launched in 2022 and immediately gained traction among developers building AI applications with complex input requirements. According to Sidhu, usage exploded in late 2023 as ChatGPT accelerated demand for apps combining documents, images, and voice.

Commercial Product Launching in Q3 2025

While Daft is already available as open source, Eventual is preparing a commercial version of the engine in Q3 2025, enabling enterprises to build full-stack AI solutions with native multimodal data ingestion.

The enterprise version will include advanced features for production deployment, monitoring, and pipeline scaling.

Eventual’s platform is currently used by notable early adopters such as:

• Amazon
  
• CloudKitchens
  
• Together AI
  

These customers span industries from logistics to food-tech to foundational AI model development—each of which depends on high-throughput unstructured data processing.

Funding and Backers: Felicis, M12, and Citi

Eventual has raised two rounds in rapid succession:

• $7.5M Seed Round (led by CRV)
  
• $20M Series A Round (led by Felicis Ventures, with participation from Microsoft’s M12 and Citi Ventures)
  

Felicis GP Astasia Myers discovered Eventual during a market analysis for multimodal infrastructure startups. She noted that Sidhu and Chia stood out as “first movers with firsthand experience,” offering unique market insight and execution strength.

“Eventual is solving the biggest bottleneck in multimodal AI infrastructure,” Myers told TechCrunch. “You need a multimodal-native engine like Daft to support where generative AI is heading.”

Industry Trend: Multimodal AI Growth Accelerates

As AI shifts from pure text to richer input formats, infrastructure that can handle multimodal data is in high demand.

According to MarketsandMarkets, the multimodal AI sector is expected to grow at a 35% CAGR from 2023 to 2028. A key driver? The fact that 90% of the world’s data is unstructured, according to IDC.

And that data is accelerating. Myers noted:

“Annual data generation is up 1,000x over the past 20 years. We’ve only just started to handle it intelligently.”

Eventual and Daft in Context