Case Study
UCU Receives Windows & ML On-Demand Environments for Machine Learning School
Ukrainian Catholic University launches their course on machine learning 90% faster with Windows & ML on-demand environments, delivered using FastGPU Stack
The Ukrainian Catholic University (UCU) is Ukraine's leading academic institution dedicated to education and research in fields as diverse as business, humanities, and sciences. The UCU's Faculty of Applied Sciences aims to help participants – undergraduates, Ph.D. students, and young professionals – proactively look into methods and tools of Machine Learning, Big Data, Computer Vision, Natural Language Processing, and Business Analytics.
The UCU was looking for a tech support provider to set up ready-to-use, on-demand Windows & ML environments for students of their ML course who needed capabilities to easily access educational content, ML/DL/CV libraries, and individual GPUs for training.
FastGPU delivered customizable Windows & ML on-demand environments on AWS VPC with an individual domain, separate EC2 instances, AMI & Docker images, ownCloud integrations, and a set of ML libraries to facilitate the launch of the ML course.
The packaged solution delivered by FastGPU allowed the participants of UCU's course on machine learning to quickly and easily access environments to complete lab work while having reduced the time needed to launch a new course by 90%.
The Faculty of Applied Sciences of the Ukrainian Catholic University is dedicated to raising awareness about Data Science. The Data Science Summer School and Machine Learning School initiatives aim to encourage everyone to learn more about AI and machine learning, as well as educates the participants about the specific methods and tools used to build, train, and enhance ML models.

Because the initiatives are launched twice a year, the faculty has to repeatedly look for a tech support provider capable to set up and deliver ready-to-use working environments. This is inefficient and endangers the course should anything go wrong.

The UCU needed on-demand environments that they could customize and use throughout the program without having to team up with third-party tech support providers. The environments as such had to be set up for machine learning, with Windows and Matlab on board, and to be easily accessible via RDP or in the browser using an HTML client. They should include all the required libraries for machine learning, deep learning, computer vision, and business analytics tasks. The participants should have access to individual GPUs to build and train ML models as required.

The course's lecturers should be able to automatically share study materials among the participants. They needed all content to be available and accessible in one single folder.
FastGPU took on a challenge to utilize the capabilities of FastGPU Stack to cover UCU's educational and research goals.

FastGPU proceeded to adapt and deploy FastGPU Stack on AWS VPC. It received an individual domain to facilitate access.

The participants' access to the study materials through FastGPU's platform was set up and enabled. To ensure smooth performance, the platform was integrated with ownCloud, which was deployed on a separate EC2 instance, with a voluminous EBS storage. A separate folder for lecturers was created; they could move educational content to that folder so that it could be instantly available to the students.

Then, an AMI was configured. It featured Windows Server, Matlab, and NICE DCV for desktop access via the browser. On top of that, a separate Docker image featuring Jupyter Notebook/Lab and other libraries, such as TensorFlow, Theano, Pandas, OpenCV, Keras, and Matplotlib was configured.

The participants were given the tools to run, stop, or delete environments created on top of a ready-to-use image and a specific EC5 C5 instance, with an automatically added storage on top. They could also run, stop, or delete Jupyter environments created on top of a Docker image and a standardized AMI image with nvidia-docker and EC2 P2/P3 instances, with an automatically supplied storage on top.

Along the way, FastGPU provided tech support to the UCU. Instances of the prepared environments, with all the required libraries, were kept track of to implement updates when required. The participants of the course could reach out to the FastGPU team for guidance.
The Ukrainian Catholic University has received a customizable solution that enables more than a hundred students of their Data Science Summer School and Machine Learning School to simultaneously study (i.e. build, train, and enhance ML models) while not overusing AWS resources. With AWS-based environments available on-demand, the lecturers can now monitor and control the amount of resources every student uses to complete a task at hand.

By using FastGPU Stack, students can now quickly access pre-setup working environments, with Windows, Matlab, and all the educational content. They no longer need to invest their time to set up the environments. They can also as easily access and take advantage of GPU environments, with ML/DL/CV libraries on board.

Thanks to FastGPU's support, the UCU has reduced the total amount of time on tech support by 90%. The university's tech personnel are no longer required to set up and administer the infrastructure for lab work. They do not have to support the course participants as well, which helps bring down the cost, too. With FastGPU's stack, the UCU aims to fast track its upcoming courses on Data Science and Machine Learning.
Save Time with FastGPU
If you are having difficulty finding the right hardware or software options, or you cannot re-configure an instance, or just do not want to spend any time on DevOps, we are here for you.
Don't hesitate to contact us for help!