GenAI Solutions Architect
Open models, tuned, optimized, and orchestrated on hardware built for the job.
Model selection, LoRA training, parameter optimization, pipelines, and the infrastructure to run it all. GENERAITR is the proof, live.
TAOS, my agent OS, running my company right now
TAOS runs GENERAITR's own operations autonomously. Live figures load from its public stats endpoint; this line is what shows if that fetch hasn't landed yet, not a stand-in number.
Picking the right model is the first decision. What happens to the footage after is the rest of the job.
GENERAITR & TAOS
Available for freelance, consulting and select collaborations. Based in Norrköping — works remotely across Europe.Creative teams don't have a generation problem, they have a repeatability problem. This is the layer between "cool demo" and "deliverable."
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.
Off-the-shelf models do what they were trained to do. I make them do what you need.
What you're actually buying
Model Selection
Open source landscape, tested and matched to the job. Not just prompted.
LoRA & Parameter Optimization
Custom trained and tuned. Real convergence data behind much of it, in Model Lab.
Pipelines
Manual, scripted, or agentic. MCP native agents when the job needs them.
Custom Tooling
Builds the tool that does not exist yet. Not just the ones that do.
Infrastructure & Hardware
Right GPU for the job. Verda, RunPod, self hosted, matched to the model.
Compliance & Provenance
C2PA signed. EU AI Act ready by default, not bolted on.
Model Lab · measured, not decorated
Empirical model testing, published where the method is reproducible: real per-step generation data, real convergence and sweet-spot detection, no eyeballed claims. Capture, transform, render — the same structure as the pipelines I build for clients, and the same audit I run on a client's pipeline before I touch it.
Start here
Book a call, or tell me what you need.
Fastest way in
Book a 30 min call→Over Google Meet. No pitch deck, just your problem and whether this is a fit.
Or connect on LinkedInNot ready to book? Tell me what you need instead.