Maksim has been writing code from a very young age. He got his bachelor and PhD degrees in Mathematics and Physics but has kept his love for computer science and programming through hobby work and high-performance numerical simulations.
Following school, he has worked at a prominent systematic hedge funds, on petabyte-size live trading systems. He has also lead teams in early stage startups as CTO. Most recently, at causaLens, Maksim and his team are building a cutting edge, autonomous, real-time system that builds adaptive predictive models on time series data.
Joining Maksim in giving this talk are Senad Ibraimoski and Kire Kolaroski.
Senad is a senior software engineer with eight years of experience developing applications across different domains -- from economics, biotechnology to VDI and ML.
He obtained his CS/CE/Mathematics degree at University of Belgrade and Tennessee Technology University. Senad is passionate in applying modern software engineering techniques to solve high-impact problems with quality always as priority.
Kire is a masters student at the Faculty for Computer Science and Engineering in Skopje, specializing in Statistical Analysis . He obtained his bachelor degree at the same university, with focus on probability, statistics and machine learning. Kire is a Junior Data Scientist at causaLens and his work consists of finding autonomous and general solutions for predictive models on time series data.
PyCon Balkan 2018 Talks
Modelling dynamical systems in Python
causaLens is a platform that automatically builds time series/dynamic system models. It is a system built on the principles of big-data processing with high performance computing in mind and its fully implemented in Python.
Most of the today's world processes can be expressed in some form of a dynamic system and time series play a big role in representing the data that is generated by the mentioned processes. These processes are difficult to deal with as underlying distributions of the data, the models and system behaviour change over time. As such online, adaptive, on-the-fly dynamic calculations are key to effective model building.
When building online models, one also has to think hard about state, especially in the high-availability setting. In this talk we will take the audience through an exploration of how we build big-data, buffering, alignment and calculation processes in Python, utilizing Numpy and Tensorflow.