Voxel51 and Bessemer Venture Partners Collaborate to Make Visual AI a Reality
A Conversation with Bessemer Investor Lindsey Li
We’re thrilled to have announced our new $30M Series B funding round, led by Bessemer Venture Partners! Bessemer is a well-established Silicon Valley venture firm with an impressive track record of backing successful startups, including Twilio, Pinterest, and Shopify, to name a few. The firm is committed to collaborating with visionary founders at the earliest stages through IPO and beyond. In this post, I’ll share why we selected Bessemer as our lead investor and dive into a Q&A with Bessemer’s Lindsey Li about why this is an exciting partnership for AI.
Why Bessemer: Building for the Long Term
While putting together this round, I was immediately struck by Bessemer’s distinctive approach. Their process stood out for its maturity, thoughtfulness, and efficiency. As I dug deeper, it also became clear that Bessemer takes a long-term perspective in the markets where they invest, including AI & ML. The team immediately started adding value to us, for example, by making several connections between our leadership and their impressive network of startup advisors and technology leaders. Equally significant was getting to know our VC partners, Lindsey Li and David Cowan, and many other exceptionally talented members of the Bessemer team. We’re proud to now join forces for the journey ahead.
Q&A: Partnering to Make Visual AI a Reality
Tell us about you, your team, and your focus at Bessemer.
Lindsey: I’ve been at Bessemer for about five years now. I started in New York and moved to San Francisco a few years ago to be at the heart of where innovation in the foundational layers of AI is happening.
Before joining Bessemer, I worked in investment banking at Goldman Sachs. There, I observed Jeremy Levine, a Bessemer partner, when Pinterest was in the process of going public (after he led the Series A). I was blown away by Jeremy’s commitment to helping leaders and companies scale through their various stages of growth. I was also drawn to Bessemer’s culture of curiosity and intellectual honesty, in which we engage and challenge each other to reach the best answer. When there was an opportunity to work with Jeremy, I jumped at it, and that was my path to venture and Bessemer.
As a venture investor, we focus predominantly on partnering with early-stage companies. We back audacious founders as early as possible in our key investment areas. We look at where technology is disrupting and revolutionizing markets and then invest in companies that have the potential to contribute bleeding-edge breakthroughs that create tremendous value in the world.
What challenges do startups face in standing out and succeeding in the crowded and noisy space of AI?
Lindsey: On the enterprise side, having spoken to dozens of technology leaders at Fortune 500 companies through this process, AI technologies often get lost in the noise of procurement, and there’s not enough time in the day to play with new products. Given these constraints, two big challenges emerge for AI technologies: how to create trust around data sharing and privacy, and how to break through the procurement paralysis.
Ways to circumvent those challenges include making software available through an open source project and cultivating an engaged community of developers around it. These are attributes we saw and liked about Voxel51’s approach in the market.
What do you think about the AI technology landscape and how it’s evolving?
Lindsey: We think about the current AI infrastructure market map in five distinct layers: compute, foundational models, data and storage, model development and training, and monitoring and observability. While there are early breakaway players in some of these categories, it’s not completely set in stone who the winners will be in each of those layers.
Additionally, if any of the five layers in the current iteration of the AI stack starts to shift, you’ll see a cascading effect across the other areas. As a result, we’ve generally shied away from any vendor that claims to be an all-in-one platform. Any software company has to be best in class in what they do at the start, build a trusted relationship with their customers, demonstrate the value their product provides, and, from that strong foundation, build out a broader product suite.
Also top of mind for us is how the AI stack evolves and differs between an AI-native organization and an AI-forward or AI-integrated organization, like a traditional enterprise. That’s why we’re excited about Voxel51’s ability to abstract away many of the complexities in building AI systems and automate previously painful manual workflows.
When Voxel51 first came across your radar, what stood out that made it worth taking a deeper look?
Lindsey: It is really rare to find a CEO with both technical acumen and commercial instincts-the ability to explain the product, the market, and the company’s vision in very clear terms. That is a unique trait we look for in the very best founders that we back.
The second reason we got excited about the opportunity is that we saw an engaged community of users. We could tell that the way the community exhibited and exuded genuine product love, which is difficult to capture and even harder to sustain.
When we talked to commercial customers spanning industries such as retail, manufacturing, robotics, and security, it became evident that Voxel51 is a serious game changer for their machine learning teams. Before Voxel51, their AI work involved heavily manual, DIY techniques. After Voxel51, their lives became easier, and they expressed how painful it would be to lose the platform. In many cases, an enterprise’s Voxel51 journey would start with one AI team and then spread organically as other teams saw the benefit it provides. Hearing customers paint the before-and-after picture and tell us that they could not conceive of going back to a world without Voxel51 is a clear signal that the company is leading the charge in a new category, a new way of doing things that is exciting to be a part of.
It was also through our customer conversations that we realized the need for a new AI orchestration layer. For example, Bosch, a leader in AI solutions from vehicle safety and autonomy to security systems to robotics, leverages Voxel51 to organize, evaluate and refine both its data and models, enabling the organization to develop robust, reliable AI applications across multiple teams and projects. My conversations highlighted how Voxel51 impacts entire organizations — developers, data scientists, management teams, and business units. It enables them to use data to inform and understand the implications of the business decisions they’re making on their AI journeys. This type of critical cross-functional usage, from AI practitioners to management teams, was eye-opening for us. We explored the parts of the AI lifecycle that precede and follow Voxel 51, specifically annotation and labeling and, later, model deployment and experiment tracking. Until Voxel51, nothing between those two filled the gap in those middle steps in the process. AI teams need this new orchestration layer to ensure that all of these workflows are happening in a way that is visible to everybody on the team and all in one place.
It’s also clear that Voxel51’s momentum aligns with the expanding market opportunity. The amount of image and video data being produced is increasing over time. People want to do something with that data; they want to clean it up, they want to use it to build models, they want to visualize it, they want to debug it. Voxel51’s vision is aligned with where we think the world is moving.
In AI, there is a mix of open and closed solutions. What are your thoughts on what’s going to be successful, and what you like about Voxel51’s open approach?
Lindsey: This is the great debate. People have different ideas of the best approach, but particularly within AI, the space overall improves with more data, more usage, more learning, and more participation. Open source excites us because it’s freely available for anyone to use, allowing people to get hands-on with the product and see for themselves the value it provides. Another big reason open source is compelling is because it provides a way to harness the power of the community in terms of product improvements.
Naturally, the company will eventually require an enterprise sales motion, but open source software gets the flywheel going quickly and efficiently.
Can you elaborate on what this investment announcement means to Voxel51’s prospects, customers, and users?
We hope it’s a resolute, positive signal about Voxel51’s potential to be core infrastructure in the world’s future with AI. I cannot understate how enlightening and powerful Voxel51’s customer feedback was to us along this journey. Customers overwhelmingly spoke of the usability of the product and clearly described how it’s made their lives better. We want to be a part of expanding and supporting Voxel51 and its vision through continued funding and partnership with the team.
Originally published at https://voxel51.com on May 16, 2024.
