Hi, I am Sebastjan
I work with Physical AI, Robotics, Linux systems and Computer Graphics.
Latest writing
The two newest essays.
The Math Behind the Magic: Markov Decision Processes in Isaac Lab
Defining a task in Isaac Lab means specifying a Markov Decision Process - states, actions, transitions, rewards, and a discount factor. Seeing the MDP turns reward design from guesswork into deliberate craft.
What Is a Policy, Really? The Thing Isaac Lab Actually Produces
A policy is the small neural network Isaac Lab trains - a function from observation to action, shipped as a checkpoint and then as portable TorchScript and ONNX files.
Latest work
A couple of recent projects.
