Thomas Gorisse

Mobile SDKs and the AI agent pipeline behind them · Creator of SceneView, open-source 3D & AR

Nantes, France

I build mobile SDKs with AI agents, and the agent systems that build them.

I work on AI agents for software delivery with Claude Code, and on cross-platform mobile SDKs. In the open, I created SceneView, the 3D & AR SDK for Jetpack Compose, SwiftUI and the web. Twenty-odd years of mobile, and still happiest turning a hard platform feature into a small API.

Try the apps GitHub LinkedIn X YouTube

An orange velvet lounge chair on wooden legs, rendered in 3D — A SceneView demo model, live in your browser. Drag to orbit.
Sheen Chair by Wayfair, CC0.

Apps

Apps you can install today

I design, build and ship these myself, on top of SceneView. They are live on Google Play and the App Store.

3D AR Model Viewer

Android · iPhone, as “SceneView: 3D & AR Viewer”

Open a 3D file from anywhere and stand it in your room at its real size.

Get it onGoogle Play Download on theApp Store
Will It Fit

Android

See a sofa, a table or a bed at its real size in your room before you buy it.

Get it onGoogle Play
SceneView Demo

Android · iPhone, as “SceneView”

Dozens of live 3D and augmented-reality scenes to try on your phone, each with the code behind it.

Get it onGoogle Play Download on theApp Store

Now

Agentic engineering, in production

Senior Mobile SDK EngineerAI agent pipeline

I build and own cross-platform mobile SDKs, and I designed the agentic pipeline behind them: how they get built, tested and shipped.

Agents implement, test and review. I own the architecture, the contracts and the quality gates, and I say no a lot.

01Multi-agent workflows
Claude Code runs the work, and each task is routed to the model and effort it actually needs.
02Independent review
The agent that writes a diff never approves it. Review is a separate agent with its own brief.
03Hard gates
Binary-compatibility checks on public APIs, screenshot tests, automated on-device QA and CI on every PR.
04Assistant-ready SDKs
MCP servers, AGENTS.md and llms.txt, so an assistant writes correct integration code on the first try.

Open source

SceneView: 3D & AR in a few lines of UI code

I created SceneView after Google stopped Sceneform. On Android it is Compose-first, on top of Filament and ARCore. On Apple platforms it is SwiftUI on RealityKit. It reaches the web, Flutter and React Native too.

1,349: stars on GitHub, live count
Over 500: public repositories declare it as a dependency

AndroidJetpack Compose · Filament · ARCore
AppleSwiftUI · RealityKit · iOS, macOS, visionOS
WebKotlin/JS · Filament.js on WebGL2
Flutter, React NativeNative bridges to both renderers

// The chair above, in a Compose screen
val engine = rememberEngine()
val modelLoader = rememberModelLoader(engine)

SceneView(engine = engine, modelLoader = modelLoader) {
    rememberModelInstance(modelLoader, "chair.glb")?.let {
        ModelNode(modelInstance = it, scaleToUnits = 1f)
    }
}

It is written for AI assistants as much as for people: the repo ships an MCP server, AGENTS.md and llms.txt. In 2026 it was accepted into Anthropic’s Claude for Open Source Program.

sceneview/sceneview Docs & live demos

SceneView is community-funded: Open Collective or GitHub Sponsors.

Work

Open source, before and around SceneView

2021 →

SceneViewcreator, maintainer

A declarative 3D & AR SDK that grew from one Android library into five platforms.

1,349
2020

Sceneform Maintainedcreator

When Google archived Sceneform, I kept it alive: Filament as the renderer, ARCore kept current, releases teams could keep shipping on. It became the base of SceneView and is now archived.

676
2021–22

Filamentcontributor

Fixes and JNI APIs upstream in Google’s real-time PBR renderer, such as transform-hierarchy access from Java, found while building SceneView on top of it.

4 merged PRs
2022–24

kotlin-mathcontributor

Quaternions from vectors, interpolation, angle helpers and epsilon comparisons for Romain Guy’s graphics math library, which SceneView depends on.

11 merged PRs

20+ years of mobile, from Java ME and BlackBerry to Jetpack Compose, AR and on-device AI. The constant has been the same job: take a hard platform feature and turn it into a small, stable API other developers can trust.

Contact