Even though admittedly inclined to breaking kitchen area appliances like ovens and microwaves, Julie Bernauer — senior answers architect for device finding out and deep mastering at NVIDIA — led the small team that productively developed Selene, the world’s fifth-swiftest supercomputer.
Adding to an previously spectacular feat, Bernauer’s staff brought up Selene as the environment went into lockdown in early 2020. They applied skeleton crews, social distancing protocols, and remote cable validation to accomplish what usually normally takes months with a greater put in team in a number of weeks.
Bernauer told NVIDIA AI Podcast host Noah Kravitz about the objective in making Selene, which was mainly to guidance NVIDIA’s researchers. Referencing her time as a doctoral university student, Bernauer clarifies how researchers are normally prevented from working on larger products thanks to price and infrastructure.
With Selene, the infrastructure is modular and can be scaled up or down dependent on what consumers demand, and will allow for distinctive sorts of research to be done concurrently. Bernauer explained that Selene is proving most helpful to autonomous vehicle and language modeling investigate at the second.
Going forward, Bernauer envisions some of the power and effectiveness of methods like Selene starting to be far more readily available on extensively obtainable gadgets, this kind of as laptops or edge merchandise these as cars.
Key Points From This Episode:
- Selene’s one of a kind, pandemic-harmless installation is even more defined in an NVIDIA site detailing the precise endeavours of Bernauer’s team and the lessons discovered from past NVIDIA supercomputers these as SATURNV and Circe.
- Bernauer joined NVIDIA in 2015, just after paying 15 years in academia. She obtained her Ph.D. in structural genomics from Université Paris-Sud, soon after which she analyzed with Nobel Prize winner Michael Levitt at Stanford.
“[Selene] is an infrastructure that people can share, exactly where we can do distinct types of research at a time” — Julie Bernauer [8:30]
“We did [Selene] for ourselves, but we also did it … to determine out how to make a merchandise better by going via the experience” — Julie Bernauer [13:27]
You Might Also Like:
Marc Hamilton, vice president of answers architecture and engineering at NVIDIA, speaks about overseeing the construction of the U.K.’s most impressive supercomputer, Cambridge-one. Built on the NVIDIA DGX SuperPOD architecture, the method will be applied by AstraZeneca, GSK, Oxford Nanopore and far more.
Hugging Facial area is far more than just an lovable emoji — it’s a organization that is demystifying AI by reworking the hottest developments in deep finding out into usable code. Research engineer Sam Shleifer talks about the company’s NLP technology, which is used at above one,000 businesses.
Bryan Catanzaro, vice president of applied deep mastering investigate at NVIDIA, walks by means of some of the newest developments at NVIDIA exploration … as well as shares a story involving Andrew Ng and cats.