Lyft Data Processing Problem: The Foundation for Future Innovation

The Rise of Eventual and the Need for Multimodal Data Processing

Sammy Sidhu and Jay Chia, the founders of Eventual, encountered a significant challenge while working as software engineers within Lyft’s autonomous vehicle program. They observed an emerging issue concerning data infrastructure, a problem anticipated to escalate with the increasing prominence of artificial intelligence.

The Data Challenge in Autonomous Vehicles

Self-driving vehicles generate substantial volumes of unstructured data, encompassing 3D scans, photographs, textual information, and audio recordings. Lyft engineers lacked a unified tool capable of simultaneously understanding and processing these diverse data types. This deficiency compelled them to assemble open-source tools, a process that proved both time-consuming and prone to inconsistencies.

According to Sidhu, Eventual’s CEO, a considerable portion – approximately 80% – of the time of highly skilled professionals, including PhDs, was dedicated to infrastructure management rather than the development of core applications. The root of these difficulties lay within the complexities of data infrastructure.

From Internal Tool to Startup Idea

Sidhu and Chia collaboratively developed an internal multimodal data processing solution for Lyft. During his subsequent job search, Sidhu repeatedly faced inquiries regarding the potential for replicating this data solution for other organizations. This recurring demand ultimately sparked the inception of Eventual.

Introducing Daft: A Python-Native Data Engine

Eventual created Daft, an open-source data processing engine built using Python. It’s engineered for rapid performance across various data modalities, including text, audio, and video. Sidhu envisions Daft as a transformative force in unstructured data infrastructure, mirroring the impact of SQL on tabular datasets.

Early Days and the ChatGPT Catalyst

The company was established in early 2022, predating the release of ChatGPT by almost a year and occurring before widespread awareness of the existing data infrastructure gap. The initial open-source version of Daft was launched in 2022, with plans to introduce an enterprise product in the third quarter of the current year.

The emergence of ChatGPT significantly amplified the demand for Eventual’s services. As more developers began constructing AI applications utilizing diverse modalities, the need for processing images, documents, and videos increased dramatically.

Expanding Beyond Autonomous Vehicles

Although Daft’s origins lie in the autonomous vehicle sector, its applicability extends to numerous other industries that handle multimodal data. These include robotics, retail technology, and healthcare. Eventual’s current clientele includes prominent companies such as Amazon, CloudKitchens, and Together AI.

Funding and Future Plans

Eventual has secured two funding rounds within a short timeframe. A $7.5 million seed round was led by CRV, followed by a $20 million Series A round spearheaded by Felicis, with participation from Microsoft’s M12 and Citi.

The recent funding will be allocated to enhancing Eventual’s open-source offerings and developing a commercial product. This product will empower customers to construct AI applications leveraging the processed data.

Felicis’s Investment Rationale

Astasia Myers, a general partner at Felicis, discovered Eventual through a market analysis focused on identifying data infrastructure capable of supporting the growing number of multimodal AI models.

Myers highlighted Eventual’s position as a first mover in this evolving space and emphasized the founders’ direct experience with the data processing challenges. She also noted the increasing significance of the problem Eventual is addressing.

Market Growth and the Importance of Multimodal AI

The multimodal AI industry is projected to experience a 35% compound annual growth rate from 2023 to 2028, as indicated by the management consulting firm MarketsandMarkets.

“Data generation has increased a thousandfold in the last two decades, with 90% of the world’s data created in the past two years,” Myers stated. “IDC reports that the majority of this data is unstructured. Daft is ideally positioned within this significant trend of generative AI being built around text, image, video, and voice – a multimodal-native data processing engine is essential.”