When I first heard the vision of Ray and Anyscale — to make distributed computing simple and more accessible — my first thought was “Yes, this was exactly what I needed 15 years ago when I was a graduate student trying to speed up my multiscale stochastic algorithms on a homegrown beowulf cluster.” My second thought was, “Surely this problem must have been solved by now, given the technology leaps that have occurred in the computing, big data and machine learning space in the last decade.” But as I dug deeper, it became apparent that this specific problem was still looking for an answer.
While there are many products that scale specific workloads (e.g. analytics, search, etc.), a universal solution to distributed computing, that is also easy to use, remains uncharted territory. Why? Because it’s incredibly hard. The simplicity and audacity of the vision, combined with the vast (and growing) need for such a universal solution, got me hooked into learning more about the team and technology at Anyscale, the creators of Ray.
Bold vision? Check. But, does the product bear this out beyond the surface? Yes, it does! And I didn’t have to look too far or deep to see that. While Ray is still in its early stages, leading organizations such as LinkedIn, Uber, Microsoft, Dow Chemical, Ant Group, and Goldman Sachs have already adopted Ray to scale various machine learning applications that power their business. Several popular machine learning libraries like Hugging Face, PyTorch Lightning, Horovod, XGBoost, and Optuna integrate with Ray to help their users easily speed up computations. And many startups are beginning to build their machine learning platforms on top of Ray.
The open source model, rapid growth and vibrant community of the Ray project reminded me of the beginning of my own journey at Elastic and how I was similarly blown away by the amazing things that Elasticsearch / ELK Stack (now called the Elastic Stack) was enabling.
Which brings me to my next reason for joining Anyscale. I know first hand the magic that is unlocked when you make powerful technology accessible and simple to use, and then directly put that into the hands of engineers. It was clear in my first few conversations with Robert, Ion and the rest of the team that Anyscale had embraced a developer-first mindset, with a relentless focus on making developers successful.
I am thrilled to join Anyscale, and I can’t wait to start telling stories of the problems we are solving, the products we are building, and the amazing outcomes our community is delivering with our technology.
New to Ray? Join the virtual Ray Summit on June 22-23 to learn more about the products, and hear directly from the community about all the amazing use cases they are enabling with Ray.