Short versionI'm a human being located in Tallinn (59.437° N, 24.754° E).
Some things I enjoy:
🤖 making machines learn
⛰ cycling, long distance running, hiking, rock climbing
📚 hunting for vintage editions at the neighbourhood book market
📖 binge reading history on Wikipedia at 1AM
🏛 discussing political science and economics over a beer
🐶 petting doggos
Back in university days I was always bewildered by how little different folks using data knew of each other. Whether it was my friends in social sciences analysing survey results with ANOVA using the IBM SPSS software or statisticians doing distributional modelling based on different assumptions or my econometrics professor who knew the proof of the Gauss-Markov theorem by heart or the AI guys claiming their deep learning was going to make the rest of science redundant, it always seemed like different faces of the same beast to me.
Whenever I asked one of these about the others, e.g. my econometrics professor about why we do not use neural networks or the AI guys about endogeneity and spurious correlations in the observational data, they couldn't tell me. This implanted a deep dissatisfaction in me, and has probably affected my career development as an empiricist and a data generalist - someone interested in all the tricks one can do using data, regardless of the domain.
I'm an applied economist by education, having done my Master's in Toulouse School of Economics. At some point in my life I was seriously considering an academic career in Economics and even got admitted to a funded PhD programme in the US, but around the same time I became interested in machine learning and its applications in the tech industry, which caused me to drop out already before starting.
After school I was lucky to receive a data lead role from Proekspert where in a year's time we established a functioning data science team that worked on a range of topics such as industrial automation, churn analytics and smart controllers. Here I learned the importance of processes, proper tooling and planning to guide the sometimes chaotic data science workflow.
Next, I worked as the head of data science for a US based real estate analytics startup. There, I drove the research roadmap revolving around everything topics such as time series forecasting at scale, portfolio optimisation (over a set of poorly defined instruments, which made off the shelf solutions not applicable) and trying to cluster neighbourhoods in a way that they would overlap with people's intuitive understanding.
Currently I'm working in a senior individual contributor role at Bolt where I've contributed to a few models running at scale in real time: dynamic pricing, fleet optimisation, routing and search experience enhancement. These days I'm mostly working on routing related stuff, making use of machine learning to capture the probabilistic nature of traffic.