In our interview series “PAIR Faces”, we ask colleagues questions about their role at PAIR Finance and get to know them personally. This time we talked to Anna Anisimova, Data Scientist at PAIR Finance.
I am from a small town in Russia surrounded by forests and lakes, 200 kilometres from St. Petersburg. Ever since I was a child, I have loved mathematics, especially solving interesting and unordinary problems. And I was also very fond of learning foreign languages. All of this led me to the field of Natural Language Processing (NLP). I moved to Berlin two and a half years ago, and I’ve been working at PAIR Finance for almost two years.
One of the most important tasks I am working on is developing an automated system for answering the most frequent German consumers’ requests in our modern debt collection process. What I like about my job is that my tasks are pretty varied: I write production code, conduct research, provide analytics for different departments, and in the process of doing so, I am constantly learning new things about the business, behavioural science and technology. I also really enjoy working with my colleagues.
Deep Learning is a branch of Artificial Intelligence (AI) inspired by the structure of the human brain. It uses artificial neural networks to learn from large amounts of raw data rather than relying on explicitly provided features. This enables deep learning models to identify complex patterns and abstractions.
In our company, Deep Learning is used for different Natural Language Processing tasks, such as identifying the language that consumers use to write emails to us or for extracting key pieces of information from a text (names, locations, companies).
Our approach to AI at PAIR Finance is business-centric. For each task, we strive to find the most optimal solution. We take into account the end-purpose of each model, the business processes of the company, potential risks and available resources. This means that we also need to be pragmatic in the tools that we use to solve the tasks: is it worth to build our own Deep Learning model or is there a pre-trained model already available open source? Should we apply a classical Machine Learning approach or use regular expressions? Eventually, what matters is the outcome and how reliably we are able to help the business.
In my free time, I write melancholic pop songs and perform on stage. In Berlin I like to attend various open mic events and get inspired by talented artists there.
Would you like to work together with Anna on the future of debt collection? Then we are looking forward to getting to know you.