Projects
Systems shipped, modules for sale.
The two flagship systems (GENERAITR and TAOS), the standalone modules they are built from (Engine, Compliance, Consistency, each sellable on its own), and Joc. Every card opens a full case page. Earlier work lives in the archive.
TAOS
A custom agent orchestration platform running 90+ autonomous agents 24/7 for a live SaaS business: a real-time constellation, not another dashboard.
GENERAITR
AI generation your team can actually repeat. Reusable templates replace the blank prompt box, and every asset ships C2PA-signed ahead of EU AI Act rules.
Since 10 Jul: a WebGL scene builder for depth/ reference conditioning.
View case→$ engine.generate(model="wan-animate", pose=openpose_ref) -> video
$
GENERAITR Engine
03Headless AI generation infrastructure built for autonomous agents, not a human at a node graph. Deterministic, model-agnostic, production-grade from day one.
$ consistency.check(clip, reference) -> drift_report
$
GENERAITR Consistency
04Headless drift detection for AI-generated media: spatial, temporal, and identity consistency, measured with the same metrics the video-generation research community uses to evaluate models.
Joc!
05A language-learning game platform. Phase 0, live now: three playable arcade minigames reactivating forgotten Catalan vocabulary, built for a 12-year-old learner. Next: Gen AI-generated media (art, audio, new vocab sets) and an agent-built content pipeline, not hand-authored levels.
$ compliance.sign(asset_id) -> c2pa_manifest + watermark
$
GENERAITR Compliance
06Automatic EU AI Act compliance for generative pipelines: every asset perceptually hashed, invisibly watermarked, and C2PA-signed the moment it's generated.