Abstract:

Technical University in Gabrovo, Bulgaria is 60 years old state university. It has three faculties - Faculty of Electrical Engineering and Electronics, Faculty of Mechanical and Precision Engineering and Faculty of Economics. Our department – Computer Systems and Technologies is part of the Faculty of Electrical Engineering and Electronics and it teaches students in computer-oriented disciplines.

As of current trends of Artificial Intelligence (AI) usage and development growth, a decision to build a specialized laboratory for study and development of AI-driven applications and solutions was made.

Artificial Intelligence is coming into our lives with many applications – starting from simple household appliances functions to complex life-saving systems. Nowadays this is one of the fastest developing scientific areas in the world. It covers wide composition of technologies, including Machine Learning (ML), Deep Learning (DL), Natural Language Processing (NLP), Computer Vision, Autonomous Systems, etc. AI development has significant impact on many industries, like healthcare, finances, transportation, robotics, security and education.

Up to the moment of writing this abstract the following activities were provided:
- Analysis of the hardware platforms available for AI applications, including Graphical Processor Units (GPUs), Tensor Processing Units (TPUs) and AI accelerators;
- Comparison of different platforms parameters, according to the needs of our laboratory (like computing power, price, energy effectiveness, etc.);
- Choice of optimal solution, according to the available funding at the present stage and development of technical specifications.
Currently public tenders for the necessary equipment are announced and the hardware delivery is expected soon, but until the full paper submission the laboratory will be built and the planned software installations will be performed and the ML and AI models will be trained and verified.

Our lab specifications include the following components:
- Server: CPU AMD Genoa 9554P with 64 Cores/ 128 threads, 256 MB Cache, 3.1 GHz; 512 GB DDR5 RAM; 16 TB SSD Storage; 2 X NVIDIA RTX5000 ADA Gen, 32GB GDDR6 GPU’s, 40 Gbit/s Ethernet card;
- 10 workstations: CPU AMD Ryzen 5 7600 6 cores, 12 threads, 3.8 GHz; 32GB DDR5-6000 RAM; Embedded AMD Radeon Graphics; 10 Gbit/s Ethernet Card; 27” IPS monitor;
- 5 NVIDIA Jetson Orin Developer Kits 32GB RAM, AI performance 200 Trillions of Operations Per Second (TOPS), 1792-core NVIDIA Ampere c GPU with 56 Tensor Cores.

As software modules we plan to use Ubuntu Linux as operating system, CUDA Toolkit for Machine Learning, along with PyTorch, TensorFlow and Keras libraries for Deep Learning, PyCharm professional for software development, MLflow and Tensorboard for AI applications and visualization.
Our research wok will include the following areas:
- Machine Learning and Deep Learning, including face recognition, generative AI models and Reinforcement Learning;
- Natural Language Processing;
- Computer Vision;
- Autonomous systems and robots;
- AI-driven telecommunication and IoT systems;
- Cybersecurity, cloud and data-center systems;
- Business process automation.

The aim of this paper is to share our experience, to describe some technical details, difficulties encountered and the way we overcome it, as well as the overall effect from implementation of the system.

Keywords:

Artificial Intelligence, Research, Laboratory, Machine Learning.