Project · Open research Proposal stage
Project S.P.H.E.R.E.: open, reproducible metrology research
Project S.P.H.E.R.E. is an open-research DeSci metrology initiative founded and architected by Michael P. Goodman. It publishes documented protocols and budgets, ships a browser-based virtual lab, and runs on a private local-AI stack that doubles as a reference architecture for regulated organizations.
By Michael P. Goodman · Last reviewed
- Role
- Founder & Principal Architect
- Organization
- SPHERE Research, LLC
- Status
- Exploratory, proposal-stage research
- Surfaces
- spheredesci.org, Virtual Lab, GitHub
Research context
Project S.P.H.E.R.E. is exploratory, open and reproducible. It is not institutionally endorsed, and nothing on this page is a scientific result or validation of the underlying hypothesis.
Project S.P.H.E.R.E. overview
The initiative is built around one rule: every claim must be reproducible by someone else. Protocols and budgets are documented in public, the simulator lets reviewers rehearse the measurement procedure before any hardware exists, and the source code and change history are open.
For the engineering side of my practice, the project is also a working proof. The same private stack that supports its tooling, with a local model, a vector database and a retrieval layer, runs with no public AI API in the data path. That is the architecture I adapt for law firms, healthcare organizations and financial institutions.
- Research site: spheredesci.org publishes the project description, scope and limitations.
- Virtual Lab: an interactive 3D simulator, covered on its own project page and live at lab.spheredesci.org.
- Reference architecture: the AI infrastructure case study documents the local LLM and private RAG build.
Project S.P.H.E.R.E. screenshots
The project's public surfaces, captured as they appear today.
Screenshots are being added
Annotated screenshots of this project will appear here. The architecture diagram below and the linked live or source pages show the current state in the meantime.
Project S.P.H.E.R.E. architecture
Four pieces make up the project, and each one is independently inspectable: documentation, simulator, source repository and the AI tooling behind them.
Design principles
- Reproducibility first. Procedures, budgets and tooling are written down so a reviewer can rerun them.
- Automatic publishing. The Virtual Lab page reads its version, notes and commits from the repository, so app updates are never hand-posted.
- Private by construction. The AI tooling runs on infrastructure the operator controls, enforced by network egress rules.
Project S.P.H.E.R.E. outcomes
What the project has produced so far. These are deliverables, not performance claims.
Open, versioned simulator
A public Virtual Lab with a maintained changelog and source repository that anyone can inspect.
Documented reference architecture
A complete write-up of the local LLM and private RAG stack, including the data boundary and the security controls.
Automated publishing pipeline
Release notes flow from GitHub to this site on a schedule, with a saved copy for search engines and no-JavaScript visitors.
Technologies used in Project S.P.H.E.R.E.
The stack splits into the simulator and the private AI infrastructure.
Simulator and web
- React
- TypeScript
- Vite
- Tailwind CSS
- Three.js
- Recharts
- GitHub Actions
- Vercel
Private AI infrastructure
- Ubuntu GPU nodes
- Ollama
- LlamaIndex
- Qdrant
- FastAPI
- React
Project S.P.H.E.R.E. FAQ
What is Project S.P.H.E.R.E.?
It is an open-research DeSci metrology initiative founded by Michael P. Goodman. It publishes documented protocols and budgets, a browser-based virtual lab and its source code. It is at the proposal stage.
Is the research institutionally endorsed?
No. The project is exploratory, open and reproducible research. It does not claim institutional endorsement or scientific validation.
How does the project relate to Michael's consulting work?
Its AI tooling is a reference build of the private local-AI architecture he deploys for regulated organizations. The case study describes that build and is not a client engagement.
Want this architecture inside your organization?
If your data cannot leave your control, the next step is an assessment of your documents, users and hardware.