Open to project work, technical leadership, and engineer-in-residence gigs. Let’s talk

Get in touch

What this is

A running log of the things I like tinkering with: writing code, electrical engineering, and mathematics. Some of it's handy, some of it's just me figuring things out in the open.

Background

My work spans multiple software layers, from consumer applications down to the systems beneath them. I began building native software for Apple platforms and the web, with a long run in user-interface engineering, then moved into low-level systems work and, more recently, applied machine learning. What ties it together is an interest in how things work down to the hardware they run on, which also keeps me close to electronics and the physics underneath.

Bookshelf

Les Misérables (Victor Hugo)

A book worth revisiting every decade. It reads differently each time.

The Feynman Lectures on Physics (Richard Feynman)

A fascinating account of how the world actually works, as much as we can figure it out anyway.

Friedrich Nietzsche

He looked at the world and wrote down what he actually saw, no filter. Hard to beat for sheer honesty.

Focus

Applied ML

Model serving, model fine tuning, agentic workflows, self-hosted deployment

Unix systems

Systems programming, lock-free concurrency, async I/O, profiling and performance

Applications

iOS and macOS, web, audio and graphics

Experience

Engineering lead

Kinoto · Porto · 2023–Present

Software developer

Berlin, London, Hamburg · 2016–2023

Application developer

Cluj, Romania · 2009–2016

Electrical CAD designer

Cluj, Romania · 2006–2009

Education

Computer Science · Philosophy

Awards

Google Android Developer Challenge

Trip Journal · Grand Prize · 2009

↓ Download CV for details

Dev log

Series · 7 parts

Back to wires: Introduction

Updated Jun 2026

A complete, ground-up course in electrical theory and code: how it's built, the shape of each lesson, and the ground it covers across eight areas from fundamentals to maintenance.

electrical
Show all 7 parts Hide parts
  1. 1. Back to wires: Introduction (above)
  2. 2. Back to wires: Direct current
  3. 3. Back to wires: DC circuit analysis
  4. 4. Back to wires: electric and magnetic fields
  5. 5. Back to wires: the electromagnetic field, light, and relativity
  6. 6. Back to wires: capacitors, inductors, and the circuits they shape
  7. 7. Back to wires: alternating current

Anthropic and the caravel problem

Jun 2026

In 1500 a country of a million people invented the ship that opened the world, got rich beyond its size, and then lost everything to latecomers with deeper pockets. Frontier AI labs are sailing the same caravel. Why the model was never the moat, and the two narrow ways out.

aiopinion

npm install and pray. Axios and TanStack, 2026.

May 2026

Two npm supply chain attacks in early 2026: the Axios compromise by Sapphire Sleet and the Mini Shai-Hulud worm by TeamPCP. What happened, how to check if you were hit, how to harden, and an honest look at how Pent would have mitigated these.

securitydevopssystems
Series · 4 parts

Vector search: The problem

Updated May 2026

Why nearest neighbor search is easy in 1D, manageable in 2D, and falls apart in 768D. The first post in a series on vector search, working from the problem itself before the algorithms.

mlalgorithms
Show all 4 parts Hide parts
  1. 1. Vector search: The problem (above)
  2. 2. Vector search: Small worlds
  3. 3. Vector search: On disk
  4. 4. Vector search: Quantization

AI and technical writing

May 2026

Almost every public article online is written with AI assistance now. So are the articles on this site. A note on how I write now, and why.

aiagentic-coding

I built a compiler frontend with Claude's help. A retrospective.

Mar 2026

A writeup of building formalang, a compiler frontend, with Claude in the loop. The actual workflow, what it cost in time and reading, where the agent helped most, and a final scorecard.

aiagentic-codingrustcompilersformalang

Self-hosting Devstral Small 2 on a 48 GB GPU

Dec 2025

Deploy Mistral's Devstral Small 2 (24B coding model) with vLLM on a 48 GB workstation GPU or rented cloud instance, then connect it to OpenAI-compatible agentic clients (Cline, Continue, OpenCode).

aidevopsself-hostingagentic-codingmistral
Series · 2 parts

The exponential thread: The puzzle

Updated Oct 2023

Starting from a 3rd-grade equation and arguing that it is the entry point to the deepest unification in modern mathematics. e, ln, sin, cos, complex exponentials, matrices, and Lie groups all turn out to be the same object viewed from different angles. The first post in a nine-part series.

math
Show all 2 parts Hide parts
  1. 1. The exponential thread: The puzzle (above)
  2. 2. The exponential thread: Why e is natural
Show 2 more Show fewer

AWS Lambda + Rust

Mar 2019

Deploying Rust on AWS Lambda in 2026: how the toolchain has changed since 2019, and what a minimal, idiomatic setup looks like today.

rustawsdevopsserverless