Who we are and what we actually do
Konpaterud is an online learning platform built around one specific subject: using deep learning to forecast prices. We keep the scope narrow because depth matters more than breadth.
Students from dozens of countries have gone through our courses — without relocating, without changing time zones, without waiting for a local university to catch up.
The background
Where this started
The platform grew out of a frustration with how machine learning is usually taught in finance contexts — lots of theory, very little contact with real data behaviour.
Price series are noisy, non-stationary, and don't behave like the clean examples in most textbooks. We built courses that deal with that directly, using actual market data as the training ground.
The model that works on backtested data often falls apart the moment the regime shifts — we spend real time on that problem.
Every lesson is designed around what a practitioner actually needs to understand, not what fits neatly into a syllabus structure.
How the platform is structured
Three things we kept in focus when building every course module and every lesson flow.
Specificity over scope
Each module targets one concrete skill — LSTM architecture, attention mechanisms, regime detection. Nothing is bundled together just to look complete.
Self-paced structure
No live sessions with fixed schedules. Students in different time zones work at their own speed, with checkpoints built in to keep progress honest.
Code-first, always
Concepts are introduced through working Python notebooks. Students see the output change as they adjust parameters, not just read about why something might matter.
We don't sell outcomes. Price forecasting with neural networks is genuinely hard, and results depend heavily on what you're modelling, how much data you have, and how you validate. The courses teach the method — what you do with it is yours to figure out.
People behind the content
Small team, specific expertise. Everyone who writes a lesson has actually worked with the methods they're teaching.
Florian Kettner
Lead instructor — deep learning architecture
Anousha Mehri
Curriculum design — time series
Søren Bloch
Platform and student support
Countries where active students are enrolled right now
Course modules covering the full forecasting pipeline
Core instructors, all with applied modelling backgrounds