What is behind our success at Strata+Hadoop World
Strata+Hadoop World becomes so large that it does not fit into Hilton conference center in London. This year, it was sold out long before the registration deadline and there were about 1800 participants from all over the world. Why it is so popular? It targets the domain where it is being decided on the success of companies from the small ones to global players such as Google, that announced during the conference that its key technology Big Table is becoming available for all developers.
Conference motto was “where big data, cutting-edge data science and new business fundamentals intersect — and merge” and this is exactly the domain which matters in today's world.
Within the conference the Startup showcase is traditionally organized. For example, the winner of this competition from 2012, the MemSQL company, received investments topping $50 mil. recently, and it is considered fastest growing database company.
This year, our company Modgen managed to win the competition http://strataconf.com/big-data-conference-uk-2015/public/schedule/detail/42407 and we would like to share some of our recipes helping us to succeed, with you.
Artificial intelligence in the cloud
In Modgen, we “think big” and prefer revolutionary reusable disruptive solutions against small incremental improvements which are also important. For many years we have been into data science and artificial intelligence research. Since 2012, we decided to capitalize our knowledge and experience. We have developed products allowing companies from several fields to improve their business.
To be specific, one of the issue that should be targeted by almost any company is churn of their existing customers. This is where we can help by predictions which customers are not satisfied and most likely to leave or stop using company services. Another example where predictive models can help is selection for marketing campaigns. Customers with highest predicted response rate are less likely to be annoyed by irrelevant offer and their loyalty seldom decreases. There are many more examples, where predictive models can be utilized. We also offer recommender systems as a service.
Our recommender systems can be used in e-commerce to generate personalized recommendations for individual users helping customers to find relevant products. This is how e-shop owners can increase their direct and indirect income and loyalty of their customers.
There are multiple companies providing recommender systems or predictive models. Our solution is unique — it is general, easy to customize and powered by artificial intelligence.
Artificial intelligence helps us anywhere, where tasks that are normally performed by people can be automated. Complexity of data science that is behind predictive models normally requires highly qualified experts with invention. However even in this area, there are many tasks that can be automated using the artificial intelligence. For example selection and parameterization of appropriate algorithms for predictive modeling or recommendation. Human experts perform this task inefficiently, because experiments evaluating predictive power of algorithms on given data set are time-consuming and there are infinite number of possibilities how these algorithms can be assembled and parametrized.
Artificial intelligence is capable of maintaining and improving thousands of algorithms in parallel for our customers so we do not need to employ army of data scientists.
Thanks to scalable cloud infrastructure and AI, we are able to adjust our system for new customer within few minutes. This is very cost efficient and our products are affordable even for smaller companies such as e-shops that would not be able to build and maintain recommender system themselves.
Moreover, according to AB test results of our bigger customers that developed recommender systems or predictive models internally, our algorithms and artificial intelligence often outperform their solutions. This is mainly due to the fact, that they run outdated or biased models and fail to maintain them as the environment and data change. Maintaining models is routine work that most of data scientists do not enjoy.
We convinced that another industrial revolution has just begun. Artificial intelligence helps people solving tasks that are net enough creative and inventive. Some tasks can do even better, because it can utilize large amount of data from different sources and increasing computational power of cloud infrastructures. There is a real danger, that half of the jobs as we know them, disappear within a few years — or more precisely will be transformed to new positions and opportunities. This is challenging for most of us. In Modgen, we have already found the way, how to cooperate with AI and we are good as a team.
Our company grows in organic way so far, we are 15 (when not taking AI into account).
Company strategy: people, products, research and innovations
Behind every success there are well educated, clever people with their own ideas that share common vision and work hard to make it happen. There are lack of such people, not only in the field of data science but also in related fields such as scalable computing, optimization or even user interfaces.
Many companies try to get such people by buying teams or drawing experienced employees from competitors. In Modgen, we build a different model, the one that we believe is sustainable and scalable. We strongly cooperate with academia, mostly with CTU in Prague https://fit.cvut.cz/en, where we participate in education of data science and artificial intelligence experts. In our joint research laboratory, students and researchers can work on innovative projects for us and our customers. In data science, it is particularly important to work with real data, design algorithms at scale and test performance of new algorithms in real traffic. This is what we bring to academia so students and researchers can work on real problems.
We use university portal http://ssp.fit.cvut.cz/en to build capacities and infrastructure that allows us to innovate our products and find new collaborators.
Originally, our plan was to focus on algorithms, their performance and automation. Later, we found out that many potential customers are not aware that predictive models or recommender systems can improve their business and have not idea in which way they should use it. Sometimes, they do not even collect data. Therefore we decided to get closer to customers to understand their needs better and help them transform their business.
Often we do not sell prediction or recommendation but develop innovative products with our strategic partners or customers. An example is a tool for editors of US publishers http://www.predictable.ly/, predicting future success of stories and recommending topics they might like and should be writing about because their target audience is most likely to read it. End users have no idea, that there are hundreds of models maintained and improved by artificial intelligence helping them to be more efficient and successful.