Suggestions
Eliot Knudsen
Director, Field Engineering at Databricks
Eliot Knudsen is a highly skilled data science professional with a diverse background in statistics, applied and computational mathematics, and machine learning. He earned his Bachelor of Science degree at Carnegie Mellon University, where he specialized in these subjects. Eliot has worked as a solutions architect at Databricks, where he uses his expertise to help clients develop effective solutions using cutting-edge technology. Eliot has also served as a Data Science Lead at Tamr and a Data Scientist at Kyruus, where he leveraged his analytical skills to drive business results. Prior to that, he worked as a Research Assistant at Carnegie Mellon and as an Administrative Assistant at the Fred Hutchinson Cancer Research Center. Eliot's unique skill set, which combines technical prowess with innovative thinking, makes him an invaluable asset to any organization seeking to harness the power of data.
Highlights
“$100 Trillion has been invested in today’s dollars into CPU-based infrastructure.”
Coatue's Philippe Laffont goes on to say that he believes that will all be ripped out to make way for a new GPU infrastructure. Link to the full interview is here: https://lnkd.in/gZP4XCG3
That’s a bit of a stretch, but directionally it is a strong argument. Microsoft, Amazon and Google are each investing $100B+ into new data center capacity - a supply-side bet on future demand of cloud services.
However, virtualization for GPUs is nascent. The last decade’s tidal wave of cloud computing was shaped largely by the vCPU and adjacent networking. With all that investment in capacity hyperscalars need to rack GPUs knowing the software to manage the infra will change!
To make matters worse, the training and inference components of an AI system have very different networking requirements. Training can go across regions and even clouds. Inference is latency bound and the rise of compound AI systems point toward co-location.
I think this transition will be slower and bigger than people expect.
After 6 1/2 years on the SA team, I'm moving to build out a new function with in Databricks: Product Adoption!
The mission to build the next Iconic Data+AI Company remains the same. The new challenge is helping customers take advantage of Databricks amazing innovation factory.
A huge thanks to my colleagues, friends and mentors in Regulated Industries who made the last 3 years building out Financial Services a blast. There are always too many people to name, but especially to Lisa Fox, Henry Hai, Christopher Robison and Bhupendra Patil for teaching me a huge amount. Also to David Totten and Arsalan Tavakoli for the opportunity to work on this new challenge.
If you're a PM with a passion for distribution or an architect with a passion for product, I am hiring product adoption architects.
