Articles
Thinking about MLOps & AI infrastructure.
Long-form writing on building ML platforms, GPU infrastructure, and the operational side of AI.
Docker for ML Engineers Who Just Want to Ship a Model
A practical Docker guide for ML engineers — writing Dockerfiles for model serving, keeping images small with multi-stage builds, handling GPUs with NVIDIA base images, and pushing to a registry.
Warning: AI GeneratedEnd-to-End ML Model Walkthrough: From model.fit() to Production
Most ML tutorials stop at training. This walkthrough covers the full loop — training a toy model, tracking experiments with MLflow, deploying to a serving endpoint, and automating retraining with scheduled GPU jobs.
Warning: AI GeneratedFive Things That Break When You Move a Model From Notebook to Production
The gap between a working notebook and a working production model is filled with subtle failures. Here are the five most common ones and how to fix them.
Warning: AI GeneratedGPU Scheduling on Kubernetes with Kueue
GPUs are expensive and shared. Kueue is a Kubernetes-native job queueing system that manages quotas, priorities, and fair sharing so teams can share a GPU cluster without stepping on each other.
Warning: AI GeneratedYour First Kubernetes CronJob for Model Retraining
A practical walkthrough of moving from manual model retraining on your laptop to an automated Kubernetes CronJob — covering container builds, secrets, and failure alerting.
Warning: AI GeneratedWhat MLflow Actually Does (and What It Doesn't)
MLflow shows up in every MLOps discussion, but what does it actually handle? A clear-eyed look at experiment tracking, the model registry, and artifact storage — plus what you'll still need to solve yourself.
Warning: AI Generated