To celebrate the Women in Tech Week, we are setting the spotlight on our AI Scientist Erlin Gulbenkoglu (LinkedIn) for this blog series. Erlin is a machine learning expert with a strong track record in building AI solutions. She has been particularly focusing on explainable AI, in other words, finding ways to provide explanations to predictions made by the model. Erlin is also an experienced speaker for both technical and business audiences, covering topics related to data privacy and machine learning model explainability. See her technical webinar on explainable AI.
Erlin holds a Master of Science in Modeling and Mathematical Methods from Paris 1 Pantheon-Sorbonne University.
Helping clients leverage their own data in a smarter way
At work Erlin likes to think that machine learning is a way to use the resources you already have in an efficient and effective way to improve your business. At Silo.AI, we are using the available resources (data) to create customized AI solutions for our clients.
“Data is a very powerful tool, but using it in a smart way makes all the difference. Each project starts by understanding the client needs and the data they have. Our job is to bring the vision for how to best utilize that data to build predictive models and to reject or not reject some hypotheses. Eventually data is able to change the client’s business”, Erlin explains.
Erlin holds learning in high respect and sees project work as a great enabler for that, both in the project team and at the client’s side. Each project strives to holistically understand the client’s business. This means gathering different points of views in order to understand the full picture.
Detecting patterns and validating anomalies
Erlin has worked on several anomaly detection cases in various industries, such as in the financial and logistics sector. Typically the project has leveraged unsupervised learning techniques. First, the target has been to recognize unusual patterns, then test if those patterns actually indicate anomalies or not.
“If you don’t know what the anomalies are in the dataset, unsupervised learning is helpful. However, after building the unsupervised models they need to be tested to verify that our models work. This can be done by using a dataset where we already know what the anomalies are.”, Erlin goes on.
Erlin has preferred to use tree-based methods to detect anomalies, as these are often easier to understand and more practical for real life scenarios. In her experience, tree-based models have outperformed more complex techniques and she has been able to make them more powerful with techniques like bagging and boosting.
Matching client’s needs with model interpretation and data privacy
As an AI expert, Erlin tries to understand what the users demand from the solutions she is creating. This approach has led her to focus on model interpretation and data privacy.
“In the past years I have observed two main concerns people have about AI development: using their personal data and trusting opaque AI systems. As an AI Scientist it is my responsibility to look for what is available in the research and apply that to the AI solutions in order to solve these issues.” At Silo.AI Erlin has been implementing the concept of Privacy by Design into the company’s AI development process.
Like other AI experts, Erlin often works with the client’s technical teams, that are in charge of explaining the problem and its value to the rest of the organization. Erlin likes to work in strong collaboration with these teams in order to utilize the client’s domain expertise and understand clearly what the client needs.
“With the client’s expertise you can find the answers you are looking for in their data much faster, which is crucial for a project with limited time”, Erlin explains.
Exploring new things wherever she goes
Originally from Turkey, Erlin has lived in five different countries. In Finland she enjoys to explore nature by hiking and likes to try out new sports such as skating and cross country skiing. At Silo.AI she’s the one organizing board game nights and movie clubs. For culture, she loves cinema and reading, and can occasionally be seen on stage with a theatre group.
Favorite Silo.AI value?
“I would say I love all the Silo.AI values but ‘Ask Why’ has been my recent favorite. Knowing my ‘why’ makes my work more fruitful, more interesting and easier. Understanding ‘why’ also helps very much within the team to collaborate effectively. I believe it is very important to allocate appropriate time for understanding your ‘why’. Then you would not waste time doing unnecessary things.”