Here in November 2022, we see a continuing onslaught of bad news: significant layoffs of incredible people as companies tighten their belts; companies that raised too much money, too fast, without core fundamentals are dying; and a changing climate where over-tightening rather than under-tightening is seemingly the new normal.
Our hearts go out to everyone adversely affected by these changes, particularly laid-off employees whose companies thought the good times would roll on and over-invested without the fundamentals to support it. Growth at all costs was a symptom of the old economy, where companies measured success based on the amount of money they raised and how fast they could acquire users at any cost rather than building scalable and enduring businesses.
At Modular, we benefit from being born in the realities of a down economy where there is no free lunch and no cheap capital. Businesses are expected to create and deliver incredible value for their customers, employees, and investors and generate healthy and sustainable revenues and margins. In 2022 and beyond, winning businesses need to be driven by “old school” first-principle business concepts rather than by “hype cycles.”
The Modular Opportunity
Practical applications of AI are proven and more are emerging all the time, but they are very expensive. These costs are driven by OpEx, CapEx, and talent with roots in the unprecedented situation where AI software is failing everyone. This software is failing because its creators never built it as a first-class product, and the teams who maintain it work in organizations without strong incentives to make it great for external customers. This software is built by a small oligopoly of “AI first” companies as a by-product of work to develop and support their AI-based products, clouds, and hardware and grow their research teams. This is why they have support teams for internal users, but not external ones. While it makes sense for these companies to prioritize their first-party products, this is a massive challenge for everyone else.
Building the next fundamental step in AI infrastructure is not “easy” or “quick.” It requires a clear vision for the future, a strong understanding of current and emerging accelerator hardware, experience deploying models into production applications, a firm grasp of the needs of AI researchers, experience building large scale software systems, a deep understanding of customer pain points, and … a sense of good design. Achieving this requires a united group of people willing to do the hard work that no one else in the industry is able (or incentivized) to address.
Building the best team in AI
In our opinion, both the opportunity and the challenge are so enormous that only a special kind of people who understand these topics and who possess a hunger to drive technology to the next level can solve them. Beyond having the most talented individuals, we believe it can only be solved if these folks actually work together with a shared vision to achieve what others - who often have much larger budgets, but much less unity - cannot. This is what drives our need to have a modern and cohesive culture, a modern approach to work, what drives us to pay among the highest in the industry, and what causes us to be so selective in creating our team.
While we are still being stealthy about our technology, product, and approach, you can be assured that the best in the industry aren’t excited about a slight increment on the status quo - we are pushing the limits of technology and product development. This includes technology that spans AI pipelines, data, models, frameworks, compilers, runtimes, serving technologies, developer tooling, and programming languages. We are customer and product first, but we believe that driving forward the fundamental technologies will bring a huge lift in the AI and developer ecosystem when we showcase it in 2023.
Work with the best
If you are interested in what we’re so focused on, please subscribe to our newsletter or apply for one of our open roles. Many Modulers have worked on ML infrastructure for years, but found themselves being held back by the politics or indecision of their previous company. Others are experts in large-scale distributed systems, serving, devops, or developer tools and don’t have great depth in AI, but they want to help reshape AI for the world. And many of us want to move past the era of LLVM to build the next generation of infrastructure systems for a future of heterogeneous compute.
The future is out there, and we are building it. 2023 will start a radical change in how people see the world of hardware and accelerated compute, from the gates up to the programming model. It is time, and the world needs this now! Follow us on Twitter and LinkedIn to learn more.