Updates:
Datapelago, backed by Eclipse Ventures, emerged from stealth this month. The most exciting moments of my career came at Cerebras Systems when we emerged from stealth and Intel Nervana when we launched the NNP training chip. When you launch ambitious technology, it is the first time you see external reaction to the idea. This is when you see the future light up! Strangely enough, being in stealth generates a curiosity that helps you hire great talent.
In the last year, we have worked with Datapelago to hire across silicon, system software, and GPU programming.
Some key hires:
1. Clement Cheung joined Datapelago to build system-level software. He previously worked on large-scale infrastructure at Meta.
2. Moises Hernandez Fernandez joined Datapelago to expand their GPU programming team to Europe. He joined from JP Morgan as VP of Applied ML and was previously at Nvidia.
3. Jimmy Aguilar Mena joined Datapelago in Europe to work with Moises on GPU programming.
Thoughts:
This month, I am taking a step back to explain our foundation.
I started Atticus to focus on building founding teams for deep tech start-ups. At the time, this seemed too narrow. On reflection, it is still too broad. The more we focus, the better our results.
Deep tech is still a relatively new term, and I am often asked what it means. This is how we define deep tech and where we focus.
Defining Deep Tech:
In 2022, I read an article by Ian Rountree at Cantos Ventures that set me on a journey to create Atticus. Ian defines deep tech:
>> Deep tech = Predominantly taking technical risk
Deep tech companies build complex systems that require a broad range of engineering disciplines, with a critical focus on co-design of hardware and software (or software with multiple layers of abstraction). This increases the technical risk and puts pressure and uncertainty on execution and scaling. It is hard enough to make the first working product, but then comes making one a month, a week, a day, and so on.
As Ian writes, you either take technical risk or market risk. Don’t take both. I am passionate about working with start-ups that build something completely new.
The interdisciplinary nature of building teams at deep tech start-ups is a fascinating challenge. It requires a deep and targeted research strategy to uncover talent from niche domains. This talent is time-consuming, expensive, and sometimes impossible to hire if big tech requires the same skill. This is increasingly common as they develop in-house custom hardware with all the software required to compile and optimize.
To put this into context, the first 100 engineers at an AI hardware start-up I worked at called Cerebras Systems were spread across 20 teams! Integrating the output of these teams into one system requires distinct leadership. Typically, founders take on this burden early on, but in doing so, they get stretched too thin and, at some point, need to hire organizational leadership.
As I read Ian’s article, I started creating the “Recruiting Case for Deep Tech”. I thought about building a firm that had lived these talent scenarios at start-ups and could come in at the seed stage to help founders build quickly and efficiently.
Where we focus:
Mapping out deep tech can get broad, but I’d consolidate it into four domains:
1. Hardware
2. Artificial Intelligence
3. Modernizing Industries
4. Techbio
Most industries defined by upfront technical risk can be placed into one of these four domains. At Atticus, we focus on the yellow boxes in our diagram below—hardware, AI, and a few verticals that build technology to modernize an industry rapidly. We started with a goal to cover all four domains, but as we got into start-ups in techbio or energy, for example, we quickly realized the talent requirement outside our expertise.
This is the map we use to define our areas of focus:
I will post an update of the map with our clients added in at the end of this year.
If you want to work at Atticus, we are hiring at the Associate and Principal/Partner levels. We are looking for people interested in the map!