Technology

Most data-extraction software separates the structure of a document from its content. But Rossum “sees” semantics and structural layout the way a human mind does. This key innovation allows Rossum’s neural networks to capture data from semi-structured documents with unrivaled precision.

Machine Learning at Scale

Rossum’s secret sauce is its in-house computer vision engine, inspired and going beyond state-of-the-art research. This core technology is complemented by our dataset – unlike any published before, comprised of tens of thousands of semi-structured documents, continuously grown and carefully annotated in detail by our data team.

Original Neural Network Research

Rossum does not just repackage past advances in text mining and natural language processing. Rossum’s architecture uses a unique approach to spatially represent textual documents and a new custom OCR engine. Both innovations are based on our own proprietary deep learning research and implemented on top of the popular machine learning frameworks TensorFlow and Keras.

In-depth Understanding of Documents

Rossum’s technology now allows a single large-scale neural network to process each document page, throughout from a pixel image to the text string output. This proves that the neural network has gained a degree of understanding of the document’s content. Rossum uses this knowledge to examine other aspects of the document using the network – from page rotation, to language, to document or page type.

Our team

Petr Baudis
Petr Baudis

Technology

Petr was one of the original authors of Git, built one of the top AIs for the board game of Go and his text understanding algorithms rival Facebook's neural networks.

Tomas Gogar
Tomas Gogar

Product

Tomas loves transforming scientific results into easy-to-use products. In his research he proposed a unique way of employing visual information in the field of text mining.

Tomas Tunys
Tomas Tunys

Science

Tomas is keen on mathematics and statistics, which drives his passion for deeper understanding of the underlying principles that make machine learning models work.

Ondrej Raska
Ondrej Raska

Business

Ondrej is co-founder and partner at Miton, CEE investment group with +100M EUR in portfolio companies. He loves the creative process of building tech company from scratch.

Marek Beran
Marek Beran

Business Director

Jana Beckova
Jana Beckova

Operations and Dataset

Elnaz Babayeva
Elnaz Babayeva

Research Programmer

Bohumir Zamecnik
Bohumir Zamecnik

AI Researcher

Simon Pavlik
Simon Pavlik

Research Programmer

Ivan Kartac
Ivan Kartac

Python Developer

Miroslav Spousta
Miroslav Spousta

Python Developer

Krystof Pilnacek
Krystof Pilnacek

Python Developer

Tomas Sorejs
Tomas Sorejs

Frontend Developer

Radek Holy
Radek Holy

Python Developer

Albert Nemec
Albert Nemec

Frontend Developer

Martin Holecek
Martin Holecek

Research Programmer

Karel Janus
Karel Janus

Business Development

Brian Kachlik
Brian Kachlik

Business Development

Antonin Hoskovec
Antonin Hoskovec

AI Researcher

Karin Fuentesova
Karin Fuentesova

Chief of AI Dataset

Jiri Pinkava
Jiri Pinkava

Python Developer

Jitka Novotna
Jitka Novotna

Python Developer

Roman Sushkov
Roman Sushkov

Research Programmer

Tomas Vesely
Tomas Vesely

Consulting Programmer

Hanka Prihodova
Hanka Prihodova

Lead Annotator

Tobias Rataj
Tobias Rataj

Partners and Alliances