About

Hi, I'm Dmitrii.

I work across Data Engineering, Data Science, Machine Learning and data platforms, with a particular interest in building systems end to end — from data collection and storage to models, APIs, monitoring and production deployment.

My background started in Data Science, but over time my work expanded into data architecture, engineering, analytics, ML systems, infrastructure, governance and technical leadership.

Dmitrii Soroka

TUMAR is my independent research and engineering project.

I created it as a place to explore ideas outside the constraints of day-to-day production work and to build complete systems around them — not just isolated models or notebooks.

The current research area is financial market microstructure: realtime market data, order book reconstruction, feature engineering, predictive modeling, execution experiments and system monitoring.

The main question is simple:

Can short-term market structure contain measurable predictive information — and does that information survive real execution constraints?

I am interested not only in model accuracy, but in the entire path from:

data → signal → execution

That means understanding where theoretical edge appears, how stable it is, what happens after fees and execution friction, and where a promising research result stops being a viable trading system.

TUMAR is also a practical engineering environment where I can experiment with realtime pipelines, data infrastructure, ML workflows, monitoring and production-style deployment.

The project will continue to evolve beyond its current focus. The broader goal is to use TUMAR as a platform for independent work in data, machine learning, quantitative research and realtime systems.

Current focus

  • Market microstructure
  • Machine learning
  • Quantitative research
  • Realtime systems
  • Data engineering
  • Production ML

Dmitrii Soroka

Data / ML / Engineering