Open to work

Eric Le · Gothenburg, Sweden

I build the layer between a language model and the job it has to do.

In practice that means agents, and the developer tools around them. Some of that was paid work, eighteen months of it at Volvo Technology, and the rest is tools I have built to make working with AI easier.

Origin

  1. 2026 shipping independently, open to the next thing

    Each project below answers a problem I ran into.

  2. jun to dec 2025 working out what to build

    Took the time to learn TypeScript and web application development, and to work out where to point AI next.

  3. jan 2024 to jun 2025 eighteen months at volvo technology

    The job was an agent that let engineering teams plot and query truck data by asking in plain language.

    It started as my master’s thesis and became a multi-step SQL pipeline with session memory.

  4. sep 2022 to jun 2024 msc in complex adaptive systems, chalmers

    How simple agents and their environment produce behavior nobody designed.

  5. sep 2019 to jun 2022 bsc in engineering physics, chalmers

    Where the modelling habits came from, before any of this was software.

Each entry made the next one possible.

Projects

aitk — terminal transcript of the install command running in a project

aitk

CLI that installs one set of agent rules, skills, and standards into every project and keeps them in step. Installs from npm as @erclx/aitk and as a Claude Code plugin.

Jobtriage, chat interface with retrieved ad cards on a React Flow canvas

Jobtriage

Live agent that triages Swedish job ads against a profile. Reachable at jobtriage.erclx.dev without a key, and the case study carries the retrieval numbers.

Caret — chat input showing a trigger-character palette of saved prompts

Caret

Chrome extension that saves prompts and drops them into Claude, Gemini, and ChatGPT from the chat input. Live on the Chrome Web Store.

diction — a scored passage reading, with the word thought flagged and the th correction shown

diction

Pronunciation trainer that scores each sound against what a native speaker actually sounds like, and reports how often it is wrong. Runs offline with nothing leaving the machine.

Looking for

what I want to build
AI tooling · LLM apps · devex
team
small to mid · close to the product
where
Sweden, Gothenburg preferred · remote
experience
two years in
terms
full-time or contract