About the talk
Our sessions have spanned research and industry sectors from materials design to banking to biology. We will provide a brief recap of these talks and the results of our panel discussions, as well as some retrospective on how we have grown SigOpt over the years to serve all of these exciting fields.
Mike leads research and engineering at SigOpt, an Intel Company. We are responsible for developing, deploying, and maintaining the SigOpt platform both by creating tools to power intelligent experimentation and making sure these tools robustly meet the usage demands of our users. He has been with SigOpt for the past 6 years, almost since its founding. Prior to joining SigOpt, he spent time in the math and computer science division at Argonne and was a visiting assistant professor at CU-Denver where he co-wrote a text on kernel-based approximation. Mike holds degrees in Applied Mathematics from Cornell and IIT.View the profile
And thank you all for attending the first-ever. They got a i in HPC Summit. I am Mike McCord head of engineering here. It's the goddamn I want to express my excitement and we can host the session like this even in a remote format. Last week, they got celebrated our 7th birthday and we have grown up a lot in that time. I've been here for six and a half of those seven years. I have always been proud of the work we have done. Here is s'got and the products that we have served to our users vision, is
to empower the world's experts. And today we heard from experts in industry and Academia that have you stick up for their own intelligence, experimentation general manager and co-founder Scott Clark, who talked about the three key elements of intelligent experimentation design Explorer optimized, a panel of the day included, an in-depth discussion of the current state-of-the-art for graph, neural networks, and I had no idea how big these crafts are nowadays. Somebody
mentioned a graph of 100 million. No, Not even being that fixed. Seems incredible to me. I'll be looking forward to the next big breakthrough for she. And his particular. Young was predicting it'll happen in search recommender systems. Maybe the truck Discovery. Maybe we'll be hearing about that big breakthrough at the next Sagat Summit. Our speakers sessions kicked off with Bondo and Salata, remind free talking about using Sagat on Transformers from prove results for their customer. We appreciate them. Staying online late into the day, all the way from
India. I hope they got some sleep. Maybe they're awake joining us right now. Diane and Alex from Accenture talked about how their work with lstms is improving predictive maintenance on oil rigs. This work is helping improve safety on these oil extraction platforms as well as helping to minimize environmental impact always grateful, we heard from Pablo at anastasiia who explains how sick out is helping his company bring time series the small and medium Enterprises throughout South America. And the rest of the world in particular.
This is an important step to making sure the smaller companies in still leverage powerful, ml tools of massive Corporation. Super time from the mental part about how his company is trying to leverage insights about Neuroscience to better design neural networks, which seems like a natural fit horse. That's where the original design of neural networks came from his goal. However is to make these neural networks as efficiently as the human brain is introducing more sparsity person as somebody who grew up doing a lot of sparse Matrix computations at Argonne National
Labs. I love to see more sparsity. Making its way into neural networks and let me take a moment here to call out that. These first four stalks were given by speakers in India, the UK, to lay and the US. I find it amazing that she got test this worldwide user base. In the middle of the day, we had a little networking session. We talked about the role that still got two plays in The Sciences weather through designing a high models model, calibration solving in Forest, problems or design spaces and effect. This would not talk about how novelis
use a cigar to conduct an efficient design of experiments. On recycling strategy is talk explained how beverage cans. Are one of the most engineer devices in the world. That's the only the space shuttle. He also explained how sick out with helping improve, that design process. I certainly appreciate all that hard work when I'm enjoying a cold beer out for a second speaker. Paul Lou from the University of Pittsburgh folk on housing, provided a more efficient alternative to genetic algorithm when searching for the preto frontier in a material design
problem in particular, he was able to additively Glass to minimize reflection, which helps the performance of devices such as solar panels, cell phones, and maybe even your computer mod. Next speaker is also working from the University, Raphael Gomes, bumper Ellie from MIT. Spoke about the work, being done to connect scientific Computing from first principles, the scientific Explorations, utilizing machine learning in particular. He discussed her multi Fidelity and experimentation gives the opportunity to learn more
quickly and I physically fabricating every model of Interest. He also spoke on how sick can be used to help solve inverse design problems in the molecular space reality. I think that this is just a really exciting topic because Addy until it's a lot of importance for compute, obviously Intel is selling a lot of resources, and I'm really excited to see all our customers successfully taking advantage of their compute resources. Our second panel explored non. Ml. Uses of Sagat. One of the exciting parts of buildings that got is it supposed
to be to work in a broad set of fields. I mentioned earlier The Sciences we had both industry and Academia speaking in this panel about how stick out can be used to design materials and structures in an efficient fashion, the cultural shift in the experimental design Community Awards. Adaptive experimentation away from the classic full factorial design. We had an interesting discussion about what action figure out and other players in the space intake, help facilitate that ship and important shift
because this design process also, often involves extremely expensive Fabrications or constitutional experiments and the more sample efficient that these can be the better. The resulting materials will be Alexandra. Johansson currently studying. Ml at Stanford spoke on the adaptation of language modeling to designing and interpreting protein. I'm not sure I'd ever heard the word proteomics before, and now I know how important it is. And again, it's just a really amazing use of AI to help improve Research In The Sciences,
specifically medicine and Therapeutics. Where the number of speakers from Intel which is no surprise because since they got to Acquisitions last year, we've been hard at work, making sure that Intel practitioners can use it to its full potential. I want to take a moment now to think everyone is until with supported our integration into the company and helped us find our footing. You have made exciting events like this Summit possible. I am just thrilled to be a part of our first Speaker from Intel cutting talked about how cigarettes help him
more efficiently soon. Deep learning recommendation model present in the MLB first Benchmark series. He also spoke about how upcoming until Hardware including the highly anticipated. Sapphire Rapids ship has been designed work efficiently on these extremely valuable. Ml. We have two speakers from Habana Labs. Talking about how Habana is working hard from within and tell to improve the ml per submissions. Azzam and Evelyn spoke on the the 1.0 of some of these submissions because the V 1.1 is secret until the
end of this month. They used to pick up for HBO in order to speed up their model experimentation process. They got platform, also gave them insights as to how hyperparameters may lead to very specific, high-performing Model Behavior. David Austin, spoke on a popular topic, for many dating site just haggle. He is a kaggle grand master and I loved hearing him get some thoughts on revisiting a competition or he was already the first place winner and improving the classification of ships
or Iceberg from very low resolution images. He described some best practices when conducting intelligent experimentation, using cigar to improve upon his own high-performing Baseline, our final Intel speaker. She and John gave a very helpful survey of the state of recommender systems. And he explained how a new end-to-end solution his group had developed, which incorporates they got helps address some of this severe costs associated with developing recommender citizens, recommender systems, for citizen data scientist, really
exciting goal and something a cigar now free from within until is definitely interested in supporting. And really all in all, it's been a great day before I became the head of engineering. I started here as sick as a research engineer. Research has been my passion for the entirety of my career. Hiccups gave me an opportunity to not just conduct research, but also put it in the hands of those who could most benefit from it to empower the world's experts so to speak.
Years ago, when we had a more private secretive group of Sagat, users wasn't always obvious that her research was actually helping. It was difficult to know whether we were making a difference. But today Today we heard from experts, regarding some of the work that stick out is supporting building out. Time, series models raft, neural networks supporting drug Discovery. Materials design, improving recycling processes. And I can't fully express my excitement regarding all of the
awesome work for sent it here today. And I also can't fully express my gratitude to the speakers for taking the time to speak about their Sagat performance. I'm so glad to hear that you're having a great time with the gaps and everyone on the research team ever won on a team. Everybody inside Sagat is incredibly excited to hear about your experiences. Thank you. Also think everyone out there. Who is Asa, got user? You are why we keep pushing to make things better. Anyone who is not a stick at user. Thank you for joining us. Today that you gave
this Summit is shot and I hope that you give cigarettes to shot. It's free forever. Maybe we'll see you up here for our next Summit. Personally, I want to thank everybody working at cigarettes for joining me on this journey. I also want to thank anyone still listening right now. You're still here after 10 hours of box. Incredible. You're the best. Thank you all for attending the first zagat's, Ai and HPC Summit. Take care.
Buy this talk
Buy this video
Our other topics
With ConferenceCast.tv, you get access to our library of the world's best conference talks.