He spoke to Econsultancy about how he came to work at EyeQuant, why a passion for understanding data is vital to his role, his favourite tools for getting the job done, and why failing early is important.
Please describe your job: What do you do?
Felix Molitor: I work as a software engineer with an academic background in cognitive science at EyeQuant. I use a combination of artificial intelligence, psychology and neuroscience to develop new ways of predicting human visual attention and perception.
Working closely with our engineering team, we use the likes of Jupyter, Keras, Tensorflow, NumPy, Pandas, React, Ansible, Terraform, AWS and Flask to build and provide models that predict the effect of visual designs on observers. We do this so that businesses can use these insights to create better user experiences.
Whereabouts do you sit within the organisation? Who do you report to?
Felix Molitor: Physically, I’m based in the Berlin office where I coordinate between engineering and R&D. Research and development, the area I currently lead, plays a crucial role for bringing new technologies and products to market. Without our longstanding partnerships with academic institutions we might be constrained by striving to be a better copy of competitors and not think creatively as to what is actually possible.
What kind of skills do you need to be effective in your role?
Felix Molitor: First, a passion for understanding data is fundamental, both in terms of its subject matter and its statistical properties. Second, experience in software engineering and architecture help to arrive at a good solution more quickly. Third, some imaginative thinking for structuring a model and optimizing its function is valuable. All these skills are eventually meant to realize a new idea, which is where I find some magic in my job.
The first time I remember getting really excited about modeling with artificial neural networks was when I was taking a course during my Cognitive Science studies at the Universität Osnabrück where we worked with a model of learning called Leabra. Today we profit from the excellent and expanding software ecosystem around Python to bring predictive modeling into practice.
Tell us about a typical working day…
Felix Molitor: … it means working together with quite a few different people, which I like a lot – there are so many different things to learn from the other members of our team. However, I also enjoy moments alone in the zone implementing a model or analyzing a data set.
What do you love about your job? What sucks?
What kind of goals do you have? What are the most useful metrics and KPIs for measuring success?
Felix Molitor: We’ve actually just completed 2020 OKR planning and my main goals are to develop projects into new products. We have planned exciting R&D initiatives with the University. And also, we want to continuously update our existing models and validate the accuracy and generalization of their predictions.
What are your favourite tools to help you to get the job done?
Felix Molitor: Apart from the libraries and frameworks mentioned earlier I was surprised a while ago that Twitter is a veritable gateway to new findings and interesting discussions in the research and software engineering areas I’m interested in. One can find “abstracts of abstracts” of new research papers and there are also live discussions on exciting topics. The short format of the service helps a bit scanning the wealth of information generated every day.
How did you end up at EyeQuant, and where might you go from here?
Felix Molitor: A friend who co-founded the company brought me on board. Interestingly, the original EyeQuant model was implemented at the University of Osnabrück. Professor Dr Peter König, Chair of Neurobiopsychology at the Institute of Cognitive Science, has been instrumental in the inception and evolution of EyeQuant.
As for departing, I recently heard of the possibility of living as a goat in Nepal. This seems like an interesting option, but currently I am very happy with my developmental perspectives at EyeQuant.
Which digital experiences have impressed you lately?
Felix Molitor: Recently, I’ve dipped into virtual reality a couple of times, and I found it quite fascinating. It still has a long way to go to have a truly immersive feeling to it for me, but there is progress in that direction, like more light and compact headsets entering the market. I imagine with techniques similar to dynamic foveated rendering, where the presented scene is adaptive to the observer’s eye movements, the visual experience itself might also become more natural.
What advice would you give someone starting out in an area like yours?
Felix Molitor: Fail early! 🙂 In the sense that it is good to rather focus on building a proof of concept than constructing an architecture of possibilities in the mind and then failing somewhere much further down the line due to a logical gap that your brain has kindly filled for you with illusion.