AIGlobalLab specializes in custom-made AI solutions. We believe that the professional implementation of artificial intelligence technologies will solve many of the challenges facing businesses.
We have developed a process that allows, on the one hand, to achieve maximum results, and, on the other hand, to optimize the cost of developing and implementing solutions.

To start a project, just send us an example of data (Table 1) and a description of the business goals that you want to achieve. We will analyze the information and offer solutions. Thus, we begin development at the presale stage.

Development process

We use a flexible development process that allows you to:

  • Take into account changes in requirements during the project
  • Maintain development in working condition
  • Synchronize customer expectations

We achieve maximum involvement of customer’s specialists in the project implementation process.


We do not invent a bicycle. Therefore, we rely on a wide range of developed algorithms, libraries, services.

Data & Text analysis

BigDL, NTLK, spaCy, Fairsec, OpenAI, GLTR, BERT, MeTA

Image processing

OpenCV, OpenVINO, PIL/Pillow, libvips, SimpleCV, Mahotas, SimpleITK, Cairo, GraphicsMagick & etc.

Objects detection and recognition

OpenCV, OpenVINO, Dlib, PyTorch, Keras, Caffe, TensorFlow, TensorFlow Lite, Torchvision, ML.NET, darknet/yolo/ImageNet

Voice recognition & audio processing

CMUSphinx, Sopare, Kaldi, VOSK, Google Voice, Microsoft Speech, Amazon Transcribe, Julius, eSpeak 

Embedded solutions

We use all the main development tools for embedded platforms: Keil, Qube MX, Atmel Studio, Atolic, IAR, Atom, Arduino, Visual Studio & more & more. We use custom development tools. Our developers have experience in creating user/kernel mode modules for Linux/Unix, MacOS, Windows Desktop/Server/IoT, *RTOS, QNX, OpenWRT, Tiny Android, mobile OS & more.

Table 1

Sample recommended data set volume for evaluation Size
Face detection 1000 frames
License plate recognition 1000 frames
Measuring the distance to the object 30 frames
Styled text generation 10Mb

Subscribe to our Digest