Our machine learning expert Markus Holopainen (LinkedIn) has been with Silo.AI since the beginning. His most well-known case is the flight delay prediction model he built for Finnair. His portfolio also includes applying machine learning to predict transit delays in e-commerce. Markus has worked on several high-profile projects in finance, including augmenting macroeconomic early warning models with machine learning and improving data quality at national banks.
Markus holds a Master of Business Administration in Information Systems and a Bachelor of Mathematics from Åbo Akademi University. He is first author of several scientific papers on augmenting conventional statistical methods for predicting financial crises with machine and ensemble learning.
Building machine learning models that work
As any AI Scientist will tell you, many pieces need to align in order to build a fully functioning machine learning solution. Its success can depend as much on the user workflow and its use case, as on the data and the chosen model. It is this multi-faceted problem that intrigues Markus in his work.
“The research and exploratory aspect of building machine learning models is fascinating. I like being able to look into vast amounts of data and to discover patterns and test hypotheses based on what I find. It’s a rewarding challenge”, Markus comments.
These patterns can reveal new insights about the client’s business and help them organize their work better. The model built for Finnair improved situational awareness at their operational control center for arriving and departing flights. In other cases, Markus’ work has helped companies identify anomalies in their transaction data, predict delays in their logistics chain and analyze vast amounts of investment documents.
Asking the right questions
With a strong background in mathematics, Markus is a problem-solver at heart. His analytical mind enjoys digging into the data and applying his skills into building something tangible. For him, understanding both the client’s business processes and the value of the model is necessary to tackle modelling problems in the best way possible.
“I need to see technology in use, applied to a business case. Data in itself contains a lot of hidden potential, but only with a clear use case the data brings out its value”, Markus explains. “Not everyone sees machine learning problems as something very concrete, but seeing how they fit into the client’s world makes them real to me.”
At Silo.AI, Markus is praised for his skills with the clients. He is fluent in explaining the model structure and output in a down-to-earth manner. He doesn’t like the idea of delivering a black box, and therefore aims for transparency by thoroughly justifying his decisions. This, in his mind, eventually helps clients to leverage the full potential of the AI solution, as they understand how it works and how it can be improved in the future.
Singer with an eye for photography
When not at work, Markus can be found on stage, either with his Turku-based quartet ‘Lök’ or choir ‘Mieskuoro Naskalit’. He is particularly fond of arranging catchy covers with a touch of humor for his quartet’s gigs. Markus is also responsible for many of the photos shared on Silo.AI’s channels as he likes (and is often asked to) bring his Sony A7 and vintage lenses to capture the moment.
Favorite Silo.AI value?
“That would be Ask Why. I believe ‘asking why’ is crucial in projects involving AI to ensure that we are doing the right things. For instance, there is no point in building advanced models based on exotic technologies if their use isn’t backed by the modelling problem and the data quantity and quality available. Similarly, if we don’t properly define the purpose of a predictive model and the criteria to measure its performance, the process becomes tough and vague going forward. Having a pragmatic approach and asking the right questions are the keys to success”, Markus explains.