Gregor Cerar, PhD

Research and Innovation Engineer | Research Fellow

Hi! I’m Gregor (Greg for short). I’m a Research and Innovation Engineer at ComSensus and a Professional Research Fellow with SensorLab at the Jožef Stefan Institute.

My work spans research software, EU-funded collaborations, and commercial projects, mostly involving backend services, data pipelines, and energy systems. Much of it is not public, so this blog reflects my personal technical interests rather than the full scope of my work.

Here I write about things I explore out of curiosity, from machine learning and time-series analysis to infrastructure and practical software. My background is in wireless communications and AI (PhD in ICT).

Outside work, I spend time with my family, hike, play chess, and play the occasional game of Uno.

LinkedIn GitHub Scholar ComSensus SensorLab

Recent posts (All posts)

Research Compute Infrastructure, Revisited: A Kubernetes Migration
11 min
compute
infrastructure
kubernetes

Three years after standing up our Kubernetes-based research cluster, preparing it for PyTorch with CUDA 13.x exposed a GPU Operator we could not bump without upgrading…

Aug 7, 2026

Finally Teaching a Network to See Without Labels
10 min
pytorch
ssl

A small CIFAR-10 study of how SimCLR, DeepCluster, and BYOL try to learn useful image representations without class labels, including the limits of one failed BYOL…

Jul 21, 2026

VQGAN+CLIP: Hallucinating Images from Text
22 min
pytorch
vqgan
clip
GAN

CLIP can score how well an image matches a text prompt, but on its own it has no way to produce one. VQGAN+CLIP closes that gap by optimizing a latent vector fed through a…

Jul 11, 2026

Diffusion Models
10 min
pytorch
diffusion

A worked implementation of a small DDPM and classifier-free guidance, with attention to the assumptions, shortcuts, and limits that affect the results.

Jul 3, 2026

Generation throughput for Qwen 3.6 35B-A3B with UD-Q4_K_XL weights as context depth increases across the runnable KV-cache and offload configurations.

Down the Rabbit Hole: Setting Up a Local LLM with llama.cpp
16 min
tech
LLM
hardware
This post describes how I turned an underused NVIDIA GeForce RTX 3090 with 24 GB of VRAM into a personal LLM endpoint. It is not a complete llama.cpp installation guide. I…
Jun 7, 2026

Reinforcement Learning: Tabular Q-Learning
7 min
code
python
mdp
reinforcement learning

This post implements tabular Q-learning for deterministic FrozenLake environments in Gymnasium. It connects the update equation to the training loop and explains what the…

Dec 17, 2025
No matching items