CEO Daniel Vila Suero tells us about Recognai’s aspirations for Data Pitch and beyond.
Describe your Data Pitch challenge idea
If data is the new petroleum, then customer data is among the most productive oil wells. Extracting value out of customer data however, is a complex and costly task. Customer data is increasingly diverse: it comes from different channels, in different formats, in various languages, with different levels of quality. Customer data management still requires enormous amounts of tedious, repetitive manpower, as well as personalised rules for every new setting, system, channel or language. What if we could teach machines to help companies these tasks? With Recognai, we can. Our solution will allow our data provider Uniserv to build, monitor and deploy robust deep learning models in multiple languages, tackling different elements of customer data management.
What does the idea set out to achieve?
Customer data management is a complex task and involves many sub-tasks, such as data cleaning and classification. Our goal is to provide a method for rapidly building and evaluating task-specific solutions based in machine learning. These solutions will increase their accuracy over time as more data becomes available and will reduce the cost of including new channels, languages, regions and systems.
What makes your idea different or unique?
Existing solutions are either based on manually-constructed rules or statistical methods. Our solution combines the robustness and adaptability of deep learning methods with the simplicity and structure of knowledge graphs.
Where did the idea come from?
The idea started with 5 years of research and its application to open data integration with several successful use cases such as the data service datos.bne.es of the National Library of Spain. When the two co-founders started working together, the technology became amplified by large-scale data processing and machine learning capabilities to provide a more powerful solution for companies.
What excites you about the challenge you applied for?
Customer data processing is a difficult task. Recent advances in deep learning methods and availability of open data, more robust solutions to many of the issues have become possible. Moreover, the idea of seamlessly combining large amounts of structured and unstructured data is very exciting as it opens new ways of solving other problems for businesses.
How did your team meet?
Two computing engineers jointly participating in a pétanque tournament in Madrid is rare, but two engineers with closely related interests, visions and complementary skills is even rarer. This is how we met. Later, we would meet in coffee shops for playing with classifiers using DBpedia while I was finishing my PhD thesis. It became clear that we had to start up Recognai together.
What’s the best thing about working with data?
It is always fascinating to see how you can combine available data with your own data in innovative ways. One example is using available book titles and descriptions at the datos.bne.es knowledge graph together with its topic classification to build a news and websites text classifier.
Anything else you want to tell us about your startup and why you do what you do?
We do what we do because we are passionate about data and language. We believe that there is space for solutions that seamlessly integrate open and structured data with language and machine learning systems.