BASEL.FM

Personal dispatches & reflections

A place where I write things down. Sometimes it's about tech, sometimes life, sometimes just something I can't stop thinking about.

Latest

Jev and the Future of Fast AI Decisions

My take on Jev: why a model built around decisions could make browser agents, tool calling, and eventually robotics faster and more practical.

29.09.20265 min read

Kalman Filter, EKF, Drones, and Robotics

Every sensor lies a little. From a plain average to the EKF that keeps drones level, robots on the map, and detection boxes on target, with five labs you can tune yourself.

14.09.202620 min read

Project Overload: The Cost of Building Faster

AI made it easier to turn every idea into a project. Daily Hub is the quiet, local command center I built to keep my work, side builds, habits, and next actions visible without turning organization into another project.

03.09.20264 min read

Branding, Guidelines, and Design Systems: The Marketing Layer Most Builders Skip

You can ship a genuinely good product and still have it look like a template. A logo, a set of guidelines, and a design system are what make a project feel considered everywhere it shows up, and that consistency is marketing you don't have to pay for twice.

24.07.20266 min read

DM Donation: Fixing How Jordan Finds Blood

Jordan needs 360,000 blood donations a year and collects about 250,000. The gap isn't generosity, it's that donors and patients never find each other in time. I built DM Donation to close that gap.

11.07.20264 min read

FDE, LLMOps, Vibe Coder: Where These Roles Actually Land

The AI job market mutated faster than anyone expected. Here's what ML Engineer, AI Engineer, FDE, LLMOps, GenAI Software Engineer, and vibe coder actually mean in practice.

01.05.20268 min read

Idea to Live in 5 Minutes

The exact workflow I use to go from an idea to a deployed, production-ready website with CI/CD, a custom domain, and traffic analytics — in minutes.

23.04.20264 min read

I Built a Qdrant Exporter in Afternoon

Qdrant had no official Prometheus exporter. I built one in an afternoon with Claude Code: Docker image, Grafana dashboard, CI/CD. Pull, run, done.

22.04.20265 min read

LiteLLM: One Gateway for Every LLM You'll Ever Run

Five LLMs, five SDKs, five different auth schemes. LiteLLM collapses all of it into one API call and gives you team caps, usage tracking, and full observability on top.

20.04.20268 min read

Serving LLMs in Secure & Air Gapped Environment

A field guide for AI Engineers who need to serve models where the internet can't follow — from staging to a truly offline GPU cluster.

09.04.202612 min read