If you struggle keeping up with the daily tasks associated with whatever you do for a living, consider the three hats that Luis Ceze wears. He’s a professor in the Paul G. Allen School of Computer Science and Engineering at the University of Washington, the co-founder and CEO of the UW machine-learning spinout OctoML, and a venture partner at Madrona Venture Group.
“It may seem like too many jobs but they are all so synergistic that it all feels natural and leads to a super fun virtuous cycle of research, productization and entrepreneurship,” Ceze said.
PREVIOUSLY: University of Washington computer science experts raise $3.9 million for machine learning startup OctoML
And he’s even busier this week, taking part at UW on Thursday in the second TVM Conference. Apache (incubating) TVM is described as “an open-source deep learning compiler stack for CPUs, GPUs, and specialized accelerators” that is intended to “close the gap between productivity-focused deep learning frameworks and performance- or efficiency-oriented hardware backends.”
Ceze and colleagues started the conference because TVM started gaining traction with academic and industrial users and they wanted a physical meeting for people to get to know each other and be even more motivated to use and contribute within the space.
“We started with a small workshop a couple of years ago,” Ceze said. “Then we did a full conference last year and 180 people flew from all over to talk about what they are doing with TVM, how they are contributing. It is great for the region because the future of machine learning is about co-designing ML models, programming models, compilers, hardware for edge and the cloud. The Pacific Northwest with UW and the industrial ecosystem here are significant players in those technology areas. Holding a conference like ours here helps bring even more attention.”
Born and raised in Sao Paulo, Brazil, Ceze has been a professor at UW for more than 12 years. He calls it more of a lifestyle than a job.
“Intellectual excitement is what gets me out of bed every day,” he said. “I love the feeling of discovering and inventing stuff, and then hopefully building something useful. And importantly, I love the teams I work with at UW and at OctoML. I have never done anything that matters alone — everything I am proud of in my life is due to other people, and I have been truly lucky to have always been surrounded by incredible people.”
He counts his life partner, Karin Strauss, among those people, calling her his intellectual and personal hero. They do research together at the Molecular Information Systems Lab at UW on using DNA for data storage, media search “and other fun things like tagging physical objects.”
When he’s not working, Ceze can be found either eating or cooking or walking/biking on Seattle’s Burke Gilman Trail.
Learn more about this week’s Geek of the Week, Luis Ceze:
What do you do, and why do you do it? I build new efficient computer systems. At UW SAMPL and OctoML we are exploring how to push the efficiency (performance, energy usage) of ML/AI to the limit by using ML/AI to optimize ML/AI systems themselves — think of Drawing Hands by MC Escher. Why? Because without significant efficiency improvements across the stack, including ML models, algorithms and hardware we limit the potential impact ML/AI can have in the world. For example, how would you make medical diagnostics systems run on a portable device (with a practical battery life)? Or large scale analytics with a reasonable budget? At OctoML we are making it very easy for users to optimizing their ML/AI models for edge, cloud and everything in between.
Now looking further out, it is clear that we can’t rely solely on conventional CMOS electronics to build future compute and storage elements — we are really close to the limits. That is why I love the work at MISL on using DNA for data storage and computation, which can bring orders of magnitude improvement in storage density and compute efficiency for tasks such as similarity search, which are very important to ML/AI.
What’s the single most important thing people should know about your field? That it is in between fields that big improvements come from. HW/SW co-design for ML/AI by using ML/AI. The intersection between biotech and computing.
Where do you find your inspiration? Art and biology. And resource limitations (e.g., energy).
What’s the one piece of technology you couldn’t live without, and why? The internet, hands down. Probably boring but true. A close second is the bicycle.
What’s your workspace like, and why does it work for you? My workspace is my laptop, my iPad, my phone and my headset. My workspace fits in my backpack :).
Your best tip or trick for managing everyday work and life. (Help us out, we need it.) List, lots of lists, including lists of lists. A good Todo app, I love Todoist. Super diligent calendar management. But most importantly, just do everything you can to get the best people around you. Hire people better than you.
Mac, Windows or Linux? Mac.
Kirk, Picard, or Janeway? Picard.
Transporter, Time Machine or Cloak of Invisibility? Time machine — you can get the other two with a time machine.
If someone gave me $1 million to launch a startup, I would … The very same one I am doing now! OctoML.
I once waited in line for … A Nintendo Wii that I only played three or four times.
Your role models: Burton Smith — a computer architect who truly knew everything about building computers, from algorithms down to people.
Greatest game in history: “Mario Bros.” SNES.
Best gadget ever: VR headset.
First computer: IBM PC with a green CRT monitor.
Current phone: iPhone 11.
Favorite app: Twitter.
Favorite cause: Just one is too hard. Sustainable technologies, universal education, universal health care.
Most important technology of 2019: AI immunotherapy.
Most important technology of 2021: Blockchain for truth.
Final words of advice for your fellow geeks: Love the people around you, be truly grateful to those that help you.
Website: UW bio
LinkedIn: Luis Ceze