Tell the Future

The Client

KÜRT Co. has grown from a small, Hungarian-owned enterprise into an international company group, developing technologies and solutions for information protection, data loss prevention and data recovery. KÜRT has earned considerable recognition and acclaim for its cutting edge technologies, including machine learning, neural networks, big data and deep learning.

For a research project funded partially by the EU, Kürt formed a consortium with two other companies, and my company became a subcontractor of one.

The Project

The research project was targeted at revealing how high-speed and real-time or near real-time log analysis using big data technologies can leverage machine learning to discover and predict anomalies. I took on multiple roles in this project from the application phase to implementation. In the first phase I was asked to review the various research materials and development concepts. Later on I was appointed to design the solution to be delivered in the course of the project. Yet, the final phase became the most interesting, as I had a chance, for first in my life, to actually do neural network and deep learning stuff.

The research project delivered the basis and proof of concept of a distributed solution with real-time monitoring of various IoT devices and other IT equipments, a centralized log collection facility for monitored data and, moreover, a centralized monitoring application using deep learning techniques to process collected data to predict upcoming failures. This has been made part of Kürt’s security incident detection and discovery solution, SeConical, and I was responsible for the machine learning and log collection components, called SeCoMaLe, standing for SeConical Machine Learning. The ecosystem used a custom communication protocol to facilitate fast and efficient data transmission and is built on an extensible architecture, enabling, for example, multiple front-ends, such as ReTiMo (as Real Time Monitoring) and one another front-end to display monitoring dashboards.

With its code developed by me, the telemetry and the central machine learning component has been built on Python, using Keras and Tensorflow for ML, whereas the ReTiMo client is a WPF application written in C#.

The Personal Side

I read my first book on AI and neural networks in around 1992. Since that I wanted to do AI. As the AI boom began soon after 2010, I wanted it more. I saw people, some of my friends, too, creating applications leveraging machine learning and deep learning. I wanted to do mine for many years without actually acting, even though they told me it’s not that hard. A professional dream came to true accidentally.

If you didn’t yet have a chance to develop an application leveraging machine learning, then I strongly suggest adding it to your bucket list. It’s superb when you put your data and an algorithm together and you see the machine will be able to tell the future. Until it works, you don’t believe it will ever do so, and then miracle just happens, sooner than expected.

Kürt and me has started to work together back in 2012. I can’t tell you what honourable it is to know they still get back to me to invite to projects time to time.