Solving real-world problems with joy for science

oddělovník1

Mathematics as a language

Spoken languages are from nature ambiguous. Mathematics is by contrast extremely clear and exact. Preceding solution with mathematical interpretation can so truly help with understanding the problem’s essence.

oddělovník2

Domain knowledge

The best algorithms and advanced methods are useless unless we have experts’ knowledge in a given field. Hence, close collaboration with our clients and a wide variety of external experts helps us to stay on top of the things.

oddělovník3

Bayesian thinking

As uncertainty has major influence on our behaviour we want to measure it precisely. The best tool for it are Bayesian methods. Only this approach enable us to combine our existing knowledge with available data and make optimal decisions.

oddělovník4

Data and Machine Learning

Human minds are unbelievably efficient machines. Yet we are not able to process huge amounts of all the available data and find their hidden patterns. Fortunately, this is exactly where the latest machine learning algorithms excel.

Our work

In Abradatas, coming up with working and elegant ideas is the most rewarding experience.

Deep
learning

Data
mining

Natural  language  processing

Cluster
analysis

Machine perception

Battery Energy Storage System

Case study:

AUTOMATIC BESS POWER MANAGEMENT

Problem

It is hard to manage power in big Battery Energy Storage System (BESS) and use it for instance for peak shaving. One has to be aware of the weather, development of electricity prices and state of the transmission grid.

Idea

Using advanced Bayesian forecasting methods (AI) based on information about the grid, weather and prices combined with smart decision making would allow us to automatically manage power in BESS, maximizing the revenue of the energy trader/user of BESS.

Our solution

We developed a communication software and power management forecasting+decision algorithms for 1 MWh (once biggest) BESS belonging to Solar Global Service in Prakšice, Czechia.

Case study:

Education Evaluation

Problem

The educational system in the Czech Republic hasn’t really changed much for a long time. Thanks to old practises students are provided with little motivation to learn and schools still have very few means to teach effectively.​

Idea

Education depends mostly on the teachers who have nor incentives nor the time to collect feedback from students. Therefore they have only small awareness of what suits their students the best.

Our solution

In cooperation with experts, we designed a questionnaire for students, then gathered answers from schools and analyze them. Finally, we provide schools with data and interactively present the results to the principal, teachers and students on Evaluace výuky web app.

Case study:

immag box

Problem

One of the important measures of yield in experimental agriculture is so-called TGW (Weight of Thousand Grains – HTS in Czech ). Manual counting is expensive, time-consuming and imprecise, but still widely used.

Idea

Using computer vision approach allows us to compute not only TGW but also other features of grain – height, width, grain damage. It is also significantly faster.

Our solution

In cooperation with IMMAG Solutions we implemented a precise computer vision algorithm for a simple camera in a controlled environment called IMMAG Box . It is able to count grains with a deviation of up to 2 % from the ground truth.

View More Case Studies
1-10467

Precision medicine

Together with SriKaalika, we are developing a methodology for a) data acquisition and storage conserving data privacy b) analytical methods for better diagnostics and treatment testing.

Modern and futuristic aesthetic black solar panels of large photovoltaic power station with wind turbines in background in nice sunny afternoon weather with partial cloudy blue sky. 3d rendering.

Trading Assistance for Solar Global Energy

We designed an algorithm and created a tool for the prediction of solar energy production. We also implemented a Bayesian decision-making system to support traders’ decisions.

angled-bays

Cheaper solutions for autonomous car parks

We developed an algorithm based on Convolutional Neural Networks
that counts arriving and departing cars.

Abradatas designed original digital image processing application that allowed us to develop an innovative approach for analysis of multiple grain features in agricultural research.
Logo IMMAG Box
Mgr. Jan Humplík Ph.D.
IMMAG Solutions s.r.o.
Abradatas solved the power management problem of our large-scale battery energy storage system by developing an automatic communication software and AI-based decision-making algorithm.
Ing. Petr Kaňovský
Solar Global Service a.s.

Our team

smolka2

František Smolka

FOUNDER

František Smolka is a long-time Czech business leader. He has gathered a wealth of experience he is now trying to hand over to Abradatas and find a way for applied mathematics into the industry. By doing so, he returns to his youthful loves – mathematics and IT.

Show more
Dostal2

Jakub Dostál

FOUNDER

Jakub Dostál is a man with luck for mentors. Ignited for mathematics, he works at the Palacký University and slowly catches up with the world of business. He currently undertakes the challenge to translate the language of science into the language of the private sector. As a student, he also took part in many international scientific competitions.

Show more
Furst_2

Tomáš Fürst

MAIN IDEOLOGIST

Tomáš Fürst is an applied mathematician who teaches at the Palacký University. He believes that mathematics is the language of natural sciences: he focuses on mathematical modelling of diverse natural phenomena and processes, processing of all kinds of data and employing the correct, i.e. Bayesian inference.

Show more