Applied Research

“Now this is not the end. It is not even the beginning of the end. But it is, perhaps, the end of the beginning.”

Sir Winston Churchill

We believe that the AI revolution is yet to come and that we are on the brink of some important qualitative leaps. Our applied research efforts are targeted towards this advance and we focus on the fields of Data Science – Knowledge Engineering and AI – Machine Learning.

The applications of our applied research are in data extraction, distributed systems, knowledge discovery, predictive analysis as well as knowledge representation. We work closely with the local technical university and with private research institutes.

Our applied research projects cover:

  • Machine learning application
  • Business intelligence and knowledge discovery
  • Natural Language Processing
  • Blockchain and smart contracts business applications

In our applied research projects our goal is to bring research and industry together and facilitate the development of novel digital solutions.

For more information about our applied research projects and capabilities Contact Us

Data Science and Knowledge Management

Data science can be regarded as the intersection between computer engineering, machine learning and traditional research.

We bring our Computer Science and Statistics expertise and, together with our partners in life science and financial domain, we develop projects for testing their hypothesis, help them develop smarter solutions and advance their digital initiatives based on real data.

Knowledge Extraction

“In God we trust, all others bring data”. The (business) world has revolved several times since the father of the quality movement, W. Edwards Deming, said that. In the wake of the fourth industrial revolution, data is available in plenty and the challenge has moved towards separating relevant data from noise, extracting the gems from the mud. Adapting the quote, we might say:

“In God we trust, all others bring actionable data”.

With this in mind, our research projects are targeted at providing actionable data to our partners through:

  • Machine learning application
  • Business intelligence and knowledge discovery
  • Natural Language Processing
  • Blockchain and smart contracts business applications

In order to address these challenges of separating relevant data from noise, together with our research and industry partners, we developed techniques for identifying relevant information using machine learning, statistic methods for isolating useful information from noise as well as NLP and domain-driven techniques for extracting information from unstructured sources.

Knowledge Representation

Our Big Data and Business Intelligence solutions are high-performance decision support systems. Our BI projects surface insights and allow our clients to identify and respond to the needs of their customers quickly.

Our knowledge representation of applied research projects are merging advanced representation models, business intelligence technologies, and domain knowledge in a way that will make data deliver more meaning and value to our partners.

Data Visualization answers to quantitative questions like: “what”, “how much”, can identify issues with data, and is good for answering a fixed number of quantitative questions regarding known correlations.

Our applied research efforts are aimed towards knowledge representation in the area of visual analytics that will help our partners to understand the data better, test their hypothesizes or discover new correlation, explore data and perform fishbone analysis without having predefined questions.

Big Data will not make you smarter but making sense of it will. Our applied research projects in knowledge representation help our partners get better insights of their data and enable them to discover new knowledge and new correlations.

Machine Learning

Through our machine learning applied research projects we aim at helping our industry partners develop innovative solutions that make them more efficient and give our research partners a go to market path.

We use both the “traditional” machine learning techniques covering the complete learning flow:

  • Preprocessing
  • Supervised, semi-supervised, unsupervised learning
  • Result assessment (defining appropriate metrics)

And the ‘deep learning’ techniques using:

  • Convolutional neural networks
  • Recurrent neural networks

Our existing projects focus on behavior pattern recognition, image recognition and predictive analysis.

Behavior Pattern Recognition

With our research partners at the Transylvanian Institute of Neuroscience (https://tins.ro), we are working on developing systems that take the results from the fundamental research at TINS to the industry. Our behavior pattern recognition systems use proprietary Neural Networks models emulating closely the behavior of the brain and offer close to real-time recognition of abnormal events with applicability in several domains from health care, like stroke prevention, to driving assistance, like taking action when the driver is somehow incapacitated or close to falling asleep.

Spatial Image Recognition

The proprietary neural networks developed by our partners at TINS, Transylvanian Institute of Neuroscience (https://tins.ro), offer an order of magnitude improvement in speed and resource consumption for image recognition. We have significant advantages over the “classical” Convolutional Neural Networks used in Deep Learning as in our applied research projects we got up to 97% accuracy in image detection while being resilient to scaling, translations and rotations.

As the model is well suited for spatial-temporal events, we build on applying this model to video and sound. Since the resource consumption is very low, adding new dimensions is easy and we work on modeling other events that can be represented in a multidimensional space.

Data Analytics

We already mentioned W. Edwards Deming, and now we go back to one of his quotes:  “Without data, you’re just another person with an opinion”. As a lot of our efforts went to data extraction we found that the more data we extract the harder it is for a human to form an opinion out of it. As Mr. Deming’s quote stands true, the reverse is also true: without an opinion, you are just another person with data.

The organizations that will be the best and the fastest at forming opinions based on their data will have a competitive edge and our applied research efforts go in this direction. We use our machine learning expertise and our industry partners’ domain knowledge to research for new applications and to enrich the data analytics and knowledge representation techniques. These techniques include dynamic, real-time analytics. They help detect more actionable insights at a higher level of granularity, improve the dashboard by presenting relevant data, and offer better predictive analysis, making our business partners more efficient and competitive.