Optimizing AI program execution in the cloud
Meet Kunal, Computer Science master student at the San Jose State University in California, who is spending his summer developing a distributed framework for Artificial Intelligence applications.
Students from many different fields join Ericsson Research for internships or thesis work. You could be next. Follow our blog to learn about the students at Ericsson Research this summer.
Hi Kunal! What do you do at Ericsson Research?
I work with Machine Learning and a new cloud platform, with the goal of optimizing how Artificial Intelligence (AI) programs execute in the cloud.
One thing I do is follow progress being made in the development of distributed systems for Artificial Intelligence. Machine Learning is currently an active area of research with promising applications in various sectors. However, Machine Learning algorithms usually require training over a large sample of data, which is a computationally heavy process.New libraries such as RAY AI execution framework is an interesting effort in this area. Ericsson Research has developed a new platform for distributed cloud computing, Cloud 3.0, which you can read about in the blog post Making Cloud easy. Based on RAY and Nefele – a component in Cloud 3.0, I work on the development of a distributed framework for AI applications. Instead of using VMs and adding one more layer of abstraction, the framework runs directly on the Linux operating system.
You can have a look at the Nefele architecture in the paper ‘Making Cloud Easy: Design Considerations and First Components of a Distributed Operating System for Cloud’published at the recent HotCloud ’18 conference.
Cloud 3.0 architecture diagram. The library Kunal is working on will be one of many applications in the Application Runtime layer.
I’ll also be implementing some performance tests to see how Cloud 3.0 performs compared to Ray running on OpenStack. Once completed, Cloud 3.0 should be able to provide a framework for Machine Learning developers to parallelize their tasks more easily on a distributed system. It’s undoubtedly a promising area of research where real benefits for the end user can be achieved.
Tell us about your education
I did my undergraduate studies at the University of Mumbai, India and I’m currently studying for my master’s degree in Computer Science at San Jose State University, California. My graduate studies focus on Artificial Intelligence and its applications. As my master’s project, I’m exploring if Generative Adversarial Networks (GAN) can be used in the area of image compression.
Why did you choose to work for Ericsson?
I was looking for industry experience in the Artificial Intelligence area. The internship at Ericsson Research will give me experience in working on various Machine Learning frameworks. This will be helpful in the future as I will understand the complete stack of how a given AI application executes. For instance, for image recognition using convolutional layer, this would mean understanding the complete process from the abstract view in which people usually work, down to the nitty gritty details of processing individual pixels and filters.
Working at Ericsson Research is especially rewarding as one gets to interact with other researchers, and whenever I’m unsure about the project or need help with the process I can contact or get a pointer to a person with relevant knowledge.
Ericsson gives me sufficient resources to work on this quite extensive assignment, and I’m given the freedom to plan and execute tasks within a broad timeline. Working at Ericsson Research is especially rewarding as one gets to interact with other researchers, and whenever I’m unsure about the project or need help with the process I can contact or get a pointer to a person with relevant knowledge.I also get to participate in initiatives such as ‘Data Science Hour’ – a forum for presenting and discussing various technologies, methodologies and concepts in this field – as well as weekly lectures on a variety of topics that take place at the Santa Clara Ericsson site. This is a workplace with people with diverse backgrounds and skillsets, and I’m sure I will gain a lot in the coming months.
What do you do when you’re not working or studying?
I like to go for the occasional trek or hike on weekends. A quiet place in nature is where I love to spend my time. I’m also interested in astronomy and I love reading non-fiction books. Sometimes I enjoy cooking, but it really depends on my mood.