My name is Jarno Lintusaari. I am a software developer with a Ph.D. in machine learning and Bayesian statistics.
In the past I have worked in e-commerce, test automation, and RPA industries. Currently I'm working with cheat detection and game security at Epic Games.
Programming has been my hobby for some 20 years, starting with a desire to make computer games and then teaching myself to code with C++.
Selected Projects
While there are many I could write about, I picked here three that I find interesting.
I recently finished a new OCR engine (extract text from an image) that I had designed for the company I was working for. It turned out to work pretty nicely being 2 times faster and 19% more accurate compared to the earlier version. The speed up was important as it enabled it be run on cloud instances without special hardware. Parts of the engine were written in C++ as some of the algorithms were too slow in Python. The engine is a mixture of image processing, deep learning, and probabilistic modelling.
Few years ago I designed and built a tailored e-commerce solution with Ruby on Rails. I even designed the outlook of it. While it's probably not the most stylish one (as I'm not really a graphic designer) it helped to achieve a 58% increase in revenue in its first fiscal year in production to roughly 0.5M EUR.
On the open source side, I was one of the core developers of the Python software ELFI for simulator based-inference based on Bayesian statistics. ELFI was published in the high-impact journal JMLR MLOSS. I designed ELFIs architecture and implemented its core functionality. That included for instance its ability to parallelize computation automatically into a computational cluster (e.g. SLURM based) which enabled generating gigabytes of data efficiently (simulator based inference is rather data hungry).
Ph.D.
I hold a Doctor of Science (Technology) degree in Information and Computer Science from Aalto University dealing with probabilistic machine learning. My doctoral dissertation titled "Steps Forward in Approximate Computational Inference" is related to a methodology called approximate Bayesian computation (also known as likelihood-free inference). It is influenced by Bayesian statistics and deals with making statistical inference by using computer simulator models instead of typical statistical models.
Blog
This blog was established to share some of my thoughts and free time projects. I'm a Finn and some posts are written in my native language.