Today, we are excited to announce that Encord has raised $30 million in Series B funding to invest fully in the future of multimodal AI data development. It’s been a little over three years since we launched our product during Y Combinator’s winter 2021 batch, where, as a two-person company, we were sitting in a California dining room, sending cold emails during the day and struggling with AI model library dependencies at night. While the company has grown significantly and evolved to meet the seismic movements in the AI space, we have not lost the two core convictions we’ve had since the early days of YC: data is the most important factor in AI and the path to building a great company is to delight customers.

Currently, the youngest AI company from Y Combinator to raise a Series B, we have grown to become the leading data development platform for multimodal AI. Our goal is to be the final AI data platform a company ever needs. We have already assisted over 200 of the world’s top AI teams, including those at Philips, Synthesia, Zeitview, Northwell Health, and Standard AI, in strengthening their data infrastructure for AI development. Our focus on creating high-quality AI data for training, fine-tuning, and validation has led to the production of better models, faster for our customers.

We’re thrilled to have Next47 lead the round, with participation from our existing investors, including CRV, Crane Venture Partners, and Y Combinator. The continued support from our existing investors is a testament to the importance of our mission and recognition that the AI department of the future is the IT department of the past.

It’s all about the data

The technological platform shift driven by the proliferation of AI will solve problems previously thought unsolvable by technology and, much like the rise of the internet in the previous generation, will touch every person on the planet. This shift has been driven by rapid innovation in the compute and model layers, two of the three main pillars for building AI applications. However, innovation, for data, arguably the most important, most proprietary, and most defensible ingredient, has been stagnant. The data layer has fallen victim to a concoction of hastily built in-house tools and ad-hoc orchestration of distributed workforces for annotation and validation, hurting data quality and ultimately model performance.

Powering the models of many of the world’s top AI teams at world-leading research labs and enterprises, we’ve witnessed firsthand the importance of having clean, traceable, governed and secure data. Knowing what data to put into your model and what data to take out is a prerequisite to true production level applications of generative and predictive AI. 

At Encord, we think and talk a lot about how we can continue to support our users in their AI adoption and acceleration journey as they cross the chasm from prototype to production. That’s why we have broken down the data problem into first principles and continue to build best-in-class solutions for each of the core components of an AI data development platform: data management & curation, annotation, and model evaluation. We seek to tie solutions to these core components together in a single, seamlessly integrated solution that works with the petabyte-scale datasets that our clients leverage in their journey to monetize their data and turn it into AI.

Some call it a data engine. Some call it a data foundry. We call it a data development platform.

The future is data is the future of us

We’re especially excited and proud of our product momentum. In the last three months alone we have added an agentic data workflow system, a consensus quality control protocol, support for audio, world-leading segmentation tracking, and many other features. We’ve also continued to make high-quality data annotation smoother and faster with the latest automation and foundation models, integrating Meta’s new Segment Anything Model into our platform less than a day after it was released and the vision-language models LLaVA and GPT4o the same week they were respectively publicly available. We plan to leverage additional capital to accelerate our product roadmap so that we can support our users—existing and new—in even more ways than we have before.

With this commitment to continued innovation of the data layer, we’re proud to publicly launch Encord Index to bring ease to multimodal data management and curation. Index is an end-to-end data management platform allowing our users to visualize, search, sort, and manage their internal data at scale. Index gives AI teams full control and governance over their data, helping them understand and operationalize large private datasets in a collaborative and secure way. It integrates seamlessly with data storage such as AWS S3, GCP Cloud Storage, Azure Blob, and others to automate the curating of the best data and remove uninformative or biased data. As a result, our customers have achieved a 35% reduction in dataset size by curating the best data, seeing upwards of 20% improvement in model performance, and saving hundreds of thousands of dollars in compute and human annotation costs. 

“Successful state-of-the-art models, like our recently released Expressive Avatar foundation model EXPRESS-1, require highly sophisticated infrastructure. Encord Index is a high-performance system for our AI data, enabling us to sort and search at any level of complexity. This supports our continuous efforts to push the boundaries of AI avatar technology and meet customer needs,” said Victor Riparbelli, Co-Founder and CEO of Synthesia, the billion-dollar generative AI company.

We’re in the early innings of building a generational company that will play a key role in the AI revolution. Thank you to our users, investors, Encordians, and partners, who make all of this possible every day. We are very excited for what’s to come.