Systems & Research
Synthetic Data Pipelines for Digital Humans
A rendering pipeline in Unreal Engine that produces labeled, multi-view video of digital humans for training models that reconstruct people from several cameras.
Custom engineering
Engineering for the parts no off-the-shelf product covers.
Off-the-shelf products stop at the edge of your workflow. The remaining piece is usually the one that touches your data, your hardware, or your engine, and it has to be typed, tested, and documented to survive handover.
For teams that need the part no product covers: a web app, an integration, a real-time service, an engine plugin, or a rendering pipeline, delivered so their own people can run it.
Custom engineering runs under every area: the web apps, integrations, and real-time servers each project needs.
Worked example
One definition, published to agents
Agent Forge is an open-source TypeScript library we wrote so a web app can offer its real operations to AI agents as typed tools. A developer defines each operation once: a stable id, a description, a JSON Schema for input, and an effect class from read to irreversible write. A consequential flag marks the actions that need a person to confirm.
The WebMCP adapter registers each capability as a tool in the browser, and an execution observer reports metadata-only lifecycle events, never input values. Agent City, WebMCP Replay Room, The Rectangle’s website, and this site all use it.
How it works
- Agent-native web apps
are the server-rendered TypeScript, Vue, and Nuxt apps on Postgres that most engagements start from.
- Agent Forge
gives any of them typed tools for agents, with consequential actions flagged.
- Spatial Venue
is the pattern for our real-time services: Rust, server-authoritative, load-tested and fuzzed.
What we can build
Web applications
Server-rendered apps in TypeScript, Vue, and Nuxt, backed by Postgres.
Integrations and APIs
Connections between the systems you already run, with retries, validation, and logs.
Real-time systems
Low-latency services in Rust and C++, such as multiplayer servers and device control.
Game engines and XR
Unity and Unreal applications, from gameplay frameworks to cross-device XR.
Rendering and synthetic data
Multi-camera rendering, ground-truth passes, and render-farm orchestration in Unreal Engine.
Cloud and deployment
Serverless and container deployments with monitoring and spending limits from the start.
Projects
Projects in this area.
Systems & Research
WebMCP Replay Room
A tennis replay workspace where a person and a browser agent review shots together through six reversible Agent Forge capabilities.
Systems & Research
GameplayFramework for Unreal Engine
An open-source Unreal Engine plugin for building replicated, data-driven games on the Gameplay Ability System. GlitchWalker and Collectors Edition are built on it.
Where to start
Start with one working feature
We build one real user flow end to end through your stack, or run an integration spike between the two systems that matter most, and write up what a full build would take.
- What you get
- Working code for one flow or one integration, with tests and a short plan for the rest.
- What you bring
- The flow or the two systems to connect
- Access to the repository, the accounts, and the data
- One person who can answer questions during the slice
Next step
Scope an engineering projectGet in touch
Tell us what you have in mind.
Describe it in a few sentences. You get a draft brief to review before anything is sent.