Superlinked Raises $9.5M in Seed Funding to Enable Enterprise-Scale AI Applications
QUICK TAKE
- Superlinked enables companies to build machine learning-powered software using their complex data faster than ever before.
- Bridging the gap between data and vector databases, Superlinked can transform everything from text to video into vector embeddings (how AI processes data) at scale.
- The company has raised $9.5 million in seed funding led by Index Ventures and Tomasz Tunguz to meet market demand and expand product capabilities.
- Superlinked was founded by a team with deep experience in machine learning and enterprise applications. Daniel Svonava spent years at YouTube building ML infrastructure and Ben Gutkovich, a former software engineer, previously supported corporations with digital transformation.
INDEX PERSPECTIVE
By Bryan Offutt, Partner at Index
When new technologies are created, they typically start with a foundational element, whether that’s the internet or databases. Following this, tools are developed to interact with and utilize the technology – message queues, caching, or APIs are such examples. Only after this, do complex applications of the new technology emerge.
However, the rapid explosion of AI into the public consciousness has created a unique situation. On the one end, we have powerful foundational Large Language Models. On the other, we have massive demand and clear visions for how to apply this technology at the application level. Yet, the intermediary layer – the advanced tools required to connect the two – is conspicuously underdeveloped.
This means it’s currently very difficult to build complex AI applications – the outcome inevitably ends up being a thin wrapper around the foundations. While great for small-scale projects, this approach falls short of meeting the demands of enterprise-level applications, which require data processing capabilities.
That’s where Superlinked comes in. The scalable solution can crunch any combination of complex and unstructured data – whether that’s text, tabular data, images, graphics, or video – into a vector embedding, the data form used by AI. Paired with a vector database, Superlinked enables machine learning to be applied to enterprise data across any industry.
What sets Superlinked apart is not just its technology but the seasoned enterprise-level experience of its founders, which is often a rarity in the AI space. Co-founder Daniel spent years at Google and YouTube building core machine learning infrastructure. It means he not only has deep technological expertise, but understands the nuances of enterprise-scale deployments.
2023 was a year of experimenting with AI applications. Now, machine learning and Large Language Models must transition into complex implementations and proven use cases. Superlinked is key to enabling this by bridging the gap between foundational AI models and practical, real-world applications.
THE DETAILS
Vector embeddings – created by converting data into a set of numbers placed within a vector space – have long been integral to the inner workings of all kinds of apps, platforms, and processes powered by machine learning. They are easy to build in simple cases, but very complicated to engineer and tune for anything more meaningful.
Addressing this with a future-proof retrieval stack entails focusing on two primary components: compute (turning data into vectors) and search (indexing and managing vectors). While considerable attention has been given to the search aspect – with over $250 million invested in vector databases over the last two years – companies have grappled with overcoming the computing challenge. This is where Superlinked plays a pivotal role. It offers a computing framework to turn any and every kind of data into vector embeddings, optimizing retrieval control, quality, and efficiency in real time so that companies can build smart software faster and easier than ever before.
“We’re very bullish on the future of vector databases, but a significant gap remains in the market. Most companies don’t have the ML and infrastructure capabilities needed to vectorize their full data landscape which limits the potential of any downstream ML application,” said Bryan Offutt, Partner at Index Ventures. “What Daniel, Ben, and the Superlinked team created bridges this gap. This is tremendously difficult to do. They’ve built a novel way for companies to turn all the unstructured and structured data they have on their users, products, or media into vectors, transforming the way companies search, analyze, and understand their data.”
Today, Superlinked announced it has received $9.5 million in seed funding. The round was co-led by Index Ventures and Tomasz Tunguz via Theory Ventures with participation from 20Sales, Firestreak, and several prominent tech executives. News of the funding comes as Superlinked launches a private alpha of its product and closes enterprise customer contracts. The fresh capital will be used to further scale to meet market demand and expand product capabilities.
The ML Compute Engine of the Future
Superlinked was founded by Daniel Svonava, an ML engineer previously with Google, where he built core ML infrastructure for YouTube Ads, focusing on the prediction of user behavior, and Ben Gutkovich a former software engineer who supported Fortune 500 corporations with digital transformation as a digital strategy consultant at McKinsey's London office.
The pair set out to build Superlinked to provide data scientists and software engineers with the ability to ship a product or feature with an ML-powered Search, a Recommender system, an Analytics pipeline, or RAG interface that works in hours or days versus the typical months or quarters by connecting the company’s data infrastructure to a vector database. They have since assembled a team with a combined 160 years of ML, software, and company building experience to accomplish this goal.
“Vectors power most of what you already do online - hailing a cab, finding a funny video, getting a date, scrolling through a shopping feed, or paying with a card. But even the best companies only use vectors for a handful of tasks - it’s just too difficult,” said Svonava, Superlinked CEO and co-founder. “We work in tandem with vector databases to put vectors at the center of enterprise data and compute infrastructure, democratizing the power that was once exclusive to a handful of tech giants.”
As enterprises start experimenting with GenAI models and upgrade their search systems to use vectors, they face the difficult choice between giving up control of the results and giving up on the promise of the technology of the future. With Superlinked, they can now have both.
Published — March 18, 2024