- In my current role, I have built up new data science function at Popsa, a printed products company. I work with both Deep Learning models in tensorflow-keras, as well as some 'good old fashioned' ML models in scipy, scikit-learn and the likes, both exposing them as APIs for mobile clients, and doing embedded processing.
I have made part of the work available here as a Python package called G&T
- In my previous position at BBOXX, I led a Data Insights team, applying machine learning methods for customer analytics - e.g. using non-linear classifiers for predicting risk of customer default
- Natural language processing - I've worked on named entity recognition for recruitment at WorkDigital, and in my spare time e.g. this sentiment analysis or this analysis of a unique spam chain
- 'Classical' computer vision - I've used a 'first principles' approach to create a pool-ball tracking model, see my pydata Warsaw talk - slides here
In my research at UCL, I was building machine learning models of human category learning.
For example, I optimised training for human learners using models which incorporated human limitations, to improve human performance in a classification task (NeurIPS paper)
My general academic interests span many areas in cognitive science and more specifically decision making.
I am also keen to apply my experience from conducting studies in business settings.
Outside of academia, I have various other interests. They include:
- Polish third sector - I was the CTO and co-founder of WayMore .
We provided a platform for people to raise money for charities through sponsored challenges. I build the web platform in Django, integrated it with payment providers, and took care of legal and financial aspect of running a limited company.
- Data analytics . I previously worked on recruitment solutions for WorkDigital .
- Film photography - @lukaszonfilm