Marty Chavez was the CIO at Goldman Sachs until he retired a couple of years ago. Of all the enterprise technology executives that I’ve worked with over the years, very few “get it” as Marty does. He’s been an entrepreneur, both as a CTO and CEO, and he’s operated billion-dollar budgets at Goldman Sachs. His insights on the evolution of technology and business are second to none. So, when Marty called to tell me how excited he was about a small start-up called Abacus.ai, my interest was piqued.
Over the past four years, there’s been an increasing amount of interest in AI and its diffusion across applications and sectors. What started as an interesting academic and research promise has turned into one of the most transformative technologies of this decade. There are many variants of AI that are now being used commercially – everything from machine learning, deep learning, reinforcement learning, and so forth. The central theme is the broad-based applicability of the technology. I’ll save the in-depth look at AI for another post, but, I’ll say this for now – at the heart of the technology is the ability to recognize patterns much in the way humans do. That is a very powerful capability that can be utilized in myriad ways from recognizing fraudulent credit card transactions, to driving autonomous vehicles, to diagnosing and curing diseases.
We’ve invested in a number of AI companies at Index – from Aurora that makes self-driving cars, to Covariant which makes the brains for robotics arms, to Scale which provides a data platform for AI applications. In addition, many of our other portfolio companies make use of AI in their applications. Not surprisingly, we’ve learned a lot from observing the efforts and developments at these companies.
One key observation is that as powerful as AI is, there are still some real challenges for the average company to wield this powerful tool. AI is still fundamentally a very "young" technology. Yes, the key concepts have been around for more than twenty years, but it was only in the last decade that sufficient computing power became available through cloud computing to make neural networks function practically. The implication of that relative technological “youth” is that AI will continue to face obstacles to broad deployment.
The first challenge is that there are not enough people with the knowledge and experience of AI. And the relatively few that do have that knowledge, have been hoarded by the big tech companies like Google, Amazon, Facebook, Apple, and Netflix. These companies not only have the financial resources to pay the salaries that AI practitioners command nowadays but also have vast troves of data that are so attractive to these individuals. Second, is that, as a technology, AI has yet to make the mainstream leap from scientific tinkering to engineering. The gaps are the professional tools that are necessary to create production-grade operations around AI, and the workflows that transform the “tweaking and tuning” of AI into a more dev ops-like process.
Undoubtedly, businesses gain a competitive advantage from the use of AI today. Yet, without the human capital or the tools that allow them to harness AI, they are stuck. It was while we were pondering through this dilemma, that Marty reached out to tell us about Abacus.ai (at the time called RealityEngines.ai). Abacus was co-founded by Bindu Reddy, Arvind Sundararajan, and Siddartha Naidu, three veterans of Google, Amazon, and Uber. Bindu had led the creation of AI verticals at AWS, Arvind was a lead engineer for Uber’s machine learning efforts in the self-driving program, while Siddartha was a founder of Big Query at Google. They have extraordinary technical depth and Bindu added the Amazon business-savvy. The Abacus Founders had made the following calculus: if the tech giants of today continue to hoard the AI talent, there is no chance for “normal” businesses to be able to build AI systems anytime in the near future. So rather than waiting for people and tools to show up, why not “outsource” the AI system to these businesses?
Deploying an ML system into production involves a number of components including data pipelines, data cleansing, re-training, model monitoring, a real-time feature store, and auto-scaling. Abacus.ai’s value proposition was to abstract all of this complexity away behind an easy to use and simple API. Their customers could deploy a large-scale AI system in production simply by pointing their dataset to Abacus’s expert AI engine. Abacus would find the right neural network architecture and, by optimizing its hyper-parameters, autonomously set up all the model infrastructure components required to deploy the model in production, re-train, model, and scale it based on prediction requests.
Within 15 minutes of our first meeting with Bindu and Arvind, we were sold. By solving these problems, Abacus harnessed the power of AI to be democratized to businesses everywhere. The only real challenge was: could they pull it off?
This is where my colleague, Bryan Offutt starred for us. Bryan has been with us at Index for a little over a year. Having concentrated his collegiate studies on AI and then put them in practice at Palantir and MemSQL before joining the VC world, Bryan brings together the perfect blend of nerdiness and insightful business perspectives on the viability of technologies in the market. We’ve come to rely on his judgment quite a bit at Index.
When I brought up Abacus at our weekly meeting, Bryan’s eyes lit up. He had been quietly attending Abacus’s tech meetups for months and had been blown away by both the technical depth of what they were doing as well as the vibrant community they had built around the company. After suggesting to Bryan that maybe he should bring these things up proactively, we jumped into the process of assessing the opportunity to invest in Abacus.
We were incredibly impressed by what we found. Abacus has assembled a world-class team of AI technologists and engineers from industry and academia. They had built the guts of their product on a shoe-string budget and were already serving a handful of (very impressed) customers with learning-based systems.
The fact that we would get to work with their legendary seed investors - Eric Schmidt, Ram Shriram, and Marty was the icing on the cake.
Today, we are proud to announce our Series A investment in Abacus.ai. As we’ve spent the last few months working with them, we are even more convinced that their vision of democratizing AI is spot on, and that they are going after a massive opportunity. Bindu, Arvind, Siddartha, along with their extraordinary team, are the right people to pull this off and the entire Index team is honored to be supporting their journey.