Skip to content

Open source

Code I publish
in the open.

Projects whose source is public: the licence, where each one stands, and what it does. Each links to a full case study with the architecture, the decisions and what is still unfinished. More of my code is on GitHub.

  • MIT53 releases since Apr 2026Independent · Creator & maintainer

    Fendix

    Application security decision platform

    Turns SAST, DAST, SCA and imported SARIF results into deterministic, confidence-scored findings and a coverage-aware release decision for CI. An open-source Go CLI (MIT) with signed releases and SBOMs, plus a commercial hosted workspace.

    Highlights

    • Confidence-scored findings from SAST, DAST and SCA, plus imported SARIF 2.1.0 from tools like CodeQL, Semgrep and Trivy.
    • Gates a release on confidence, not severity alone, and every point of a finding's score has a written reason.
    • Coverage-aware: a scanner that didn't run is recorded as skipped or failed, never mistaken for a clean pass.
    • Deterministic: no machine-learning model in detection, scoring or decisions.
    • Signed releases with CycloneDX and SPDX SBOMs and SLSA provenance, and no telemetry from the CLI.
    • Go
    • Python
    • Semgrep
    • Django
    • Next.js
    • Sigstore cosign
    Source on GitHub Case study Website
  • Apache-2.0Working prototypeIndependent · Creator & maintainer

    Qatf

    Long videos into vertical clips, Arabic first

    Open-source tool that turns a long video into 9:16 clips with word-by-word captions for Reels, Shorts and TikTok. Whisper transcribes, a swappable language model picks the moments, and every cut is snapped onto a real word boundary. Arabic right-to-left captions are a first-class path. CLI, FastAPI service and React web UI; a working prototype.

    Highlights

    • Whisper transcribes, a language model picks the moments, and every cut is snapped onto a real word boundary.
    • Word-by-word captions on right-to-left Arabic as well as English, in four styles, checked on rendered frames.
    • Any model in one setting: Anthropic, OpenAI, Kimi, GLM, OpenRouter, or local Ollama and vLLM.
    • Review before rendering: edit the picks, titles and order; edited cuts are moved back onto words.
    • CLI, HTTP API and an English and Arabic web UI, started with one docker compose up.
    • Python
    • FastAPI
    • faster-whisper
    • ffmpeg
    • React
    • TypeScript
    • SQLite
    • Docker
    Source on GitHub Case study Feature list