nova-agent - machine learning

algorithms taught by practitioners

nova-agent builds structured masterclasses around machine learning algorithms - not survey courses, but deep technical instruction from researchers and engineers who use these methods daily. Research, monitoring, and automation are the backbone of what we teach.

18+ algorithm modules

6k enrolled learners

4.1 avg. rating

Instructor explaining a gradient descent algorithm on a whiteboard

Step-by-step algorithm breakdowns - theory grounded in working code

An AI for Mac that you can entrust with real tasks

Nova handles research, browser work, and automation so you can stay focused on decisions that actually require your judgment. That same philosophy shapes every course we build - reduce friction, increase depth.

Our curriculum grew from a single observation: most ML courses explain what an algorithm does, but skip the reasoning behind parameter choices, failure modes, and real deployment constraints.

What separates this platform from a typical course library

Each module centers on a single algorithm - decision trees, gradient boosting, attention mechanisms - with worked examples using real datasets, not synthetic toy problems.
Instructors walk through failure cases explicitly - what breaks, why it breaks, and what diagnostic steps surface the problem before it reaches production.
Content aligns with regional educational standards in the La Grange area and connects learners to local initiatives in applied data science and civic technology.

The people who build the curriculum

Three instructors, each with a distinct technical background. They disagree on tooling choices and that tension shows up in the material - which is the point.

Probabilistic models - Bayesian inference and uncertainty quantification in practice

Teodora Vašíčková

Probabilistic Methods - Lead

Spent eight years at a computational biology lab before moving into applied ML education. Teaches Bayesian networks and Gaussian processes with an emphasis on what the math actually assumes.

Ensemble methods - when stacking models compounds errors instead of reducing them

Rémi Aubert

Ensemble - Methods

Worked on fraud detection pipelines for a mid-size fintech before joining nova-agent. His modules on gradient boosting cover the hyperparameter interactions most documentation glosses over.

Deep learning architectures - attention, transformers, and when simpler models still win

Sigríður Björk

Deep Learning - Architectures

Researched sequence modeling at a university NLP lab. Brings a critical view of transformer hype - her modules spend as much time on when not to use attention as on how it works.

Nova-agent team collaborating on curriculum design at a shared desk

How the curriculum gets reviewed

Every module goes through a structured peer review before publication. One instructor teaches it, a second checks the technical claims, and a third verifies the exercises produce the stated outcomes on a clean environment.

Modules are updated when a dependency changes or when learner feedback surfaces a consistent gap. There is no fixed release schedule - only a threshold for accuracy.

Questions about the program?

712 E Elm Ave, La Grange, IL 60525 - [email protected]

view the program get in touch