Auxance · Hub
EN ⇄ FR

Full-stack student at Holberton building production systems by day, training myself…

WhatI'mdoingrightnow.

LEARNING

AI Journey — word embeddings, by hand

Deep in the representations chapter of AI Journey: how words become vectors — co-occurrence, word2vec, skip-gram, negative sampling — implemented in NumPy before any framework touches it. The rule stands: never train a model I can’t explain.

BUILDING

JobMatch — Demoday, July 2026

Shipping the matching platform for Demoday: local CV extraction with Ollama, offers aggregated from four sources, an explainable 0–100 score.

WRITING

The book — written here

Writing the road toward ML research as a real book, chapter by chapter, directly inside this portfolio.

Selectedwork.

  1. 01AI JourneyLearning

    The public log of a transition from systems programming to AI — math, ML fundamentals and end-to-end projects, with one rule: never train a model I can’t explain.

  2. 02JobMatchBuilding

    A CV ↔ job-offer matching platform: a local LLM reads the résumé, then every aggregated offer gets an explainable 0–100 compatibility score.

  3. 03ManagerShipped

    Count the days, keep the log — a local-first desktop app for discipline: day counter, focus timer, planner, journal. Free forever, no account, no tracking.

  4. 04StelaShipped

    A minimalist, glassy Markdown notes app for Windows — plain .md files synced straight to your own Google Drive. No server, no tracking, no lock-in.

All code · GitHub ↗

Let'stalk.

Open to ML research collaborations, and quiet conversations about the path

[email protected]

Follow