Design data from practicing architects.
A generated floor plan can look convincing and still have no way out. We commission vetted architects to create, label and critique the design data AI models learn from.
Architecture, with proof.
Generative design and spatial reasoning models produce plans that fail basic architectural sense: a stair that arrives nowhere, doors swinging into each other, a bedroom with no window, rooms out of scale, or circulation that ignores egress. HUMXN commissions credential-checked architects and architectural designers to label plans and drawings, build models, critique generated designs and write the spatial reasoning that teaches models how buildings actually work.
Drawings, models and renderings are sourced from consenting creators who hold the rights, or made new for the request, with consent bound to each item. Images are fingerprinted with perceptual hashes and models with 3D shape fingerprints, then checked for duplicates. Human-origin checks read embedded metadata, label provenance records who made and confirmed each label, and a second architect reviews every item before delivery.
For architects, this is paid, remote work that draws on your design training and professional judgment. You annotate plans, critique generated designs and review a colleague's work, at a rate shown before you accept each job. The work creates data for AI systems. It is not design services for any client, and you do not stamp or certify anything. Approved work is paid automatically through Stripe.
Built to your specification.
Every engagement starts from a written spec and a pilot batch. These are the most common requests we source for architecture.
Annotated floor plans
Plans labeled for rooms, walls, openings, circulation, dimensions and use, following a written schema.
Drawing sets
Plans, sections and elevations from consenting creators or drawn new, with consistent labeling across sheets.
BIM and 3D models
Building models with elements classified and properties recorded, fingerprinted and provenance-signed.
Generated design critique
Architects review AI-generated plans for egress, accessibility, daylight, structure and buildability.
Spatial reasoning items
Questions and worked answers about layouts, adjacency, scale and circulation for evaluating multimodal models.
Image and rendering labels
Architectural photography and renderings described by style, material, element and spatial quality.
01Plausible but unusable plans
Generated layouts often break basic rules. Architects judge them as a reviewer would, not only on appearance.
02Code awareness
Egress and accessibility depend on jurisdiction. Labels note which assumptions apply so models do not generalize wrongly.
03Rights in design work
Drawings and models are creative work. Consent bound to each item and fingerprints keep the source traceable.
Paid work in your field.
Remote and flexible, with the rate stated before you accept. Every job is reviewed by a second expert, and approved work is paid automatically.
- A professional degree in architecture, such as a BArch or MArch
- Architectural registration or license, for review and critique tasks
- Several years of practice in residential, commercial or institutional work
- Fluency in BIM and CAD tools
- Specialisms such as accessibility, housing, healthcare or interiors
Annotate plans and drawings
Label rooms, openings, circulation and dimensions on plans and sections to a written schema.
Critique generated designs
Review AI-generated layouts for egress, accessibility, daylight and buildability, with written notes.
Write spatial reasoning items
Pose questions about layouts and adjacency and write the correct, reasoned answers.
Review another architect's labels
Act as second reviewer on a colleague's annotations and critiques before delivery.
Architecture: asked often.
What architectural data can you provide?
Annotated floor plans and drawing sets, BIM and 3D models, critiques of generated designs, spatial reasoning questions and labeled renderings and photography. Requests are scoped by building type, region and format, then matched to architects verified in that area. A pilot batch confirms the schema before scaling.
Where do the drawings and models come from?
From consenting creators who hold the rights, with consent bound to each item, or drawn new by architects for your request. Every item is fingerprinted at intake, checked for duplicates and screened for human origin using embedded metadata, and carries signed provenance.
Do I need to be a licensed architect?
Not for every job. Plan annotation and labeling are often open to architectural designers with a professional degree and practice experience. Critique and code-related tasks usually ask for registration. The work is data creation for AI, not design services for a client, and you do not stamp or certify anything.
How do I get paid?
Rates are set by field and task and shown before you accept. When a second architect approves your work, it is credited to your balance and paid automatically through Stripe after your identity and tax details are verified. Each job is itemized in your ledger.
How much time is required?
There is no minimum commitment. Opportunities arrive when they match your verified profile, each with a stated scope and deadline, and you choose which to accept. The work is remote and fits around practice or teaching, so you can take more work between projects and less near a deadline.
Can we license plans exclusively?
Yes. Commissioned drawings, models and labels can be licensed exclusively, with scope, term and exclusivity agreed in writing. Images can be delivered as COCO or WebDataset and metadata as JSONL or Croissant, each with a signed receipt listing every item by ID and hash. Standing orders can add new building types over time.
Architecture, done by experts.
Need expert-made data in this field, or have the expertise to make it? Start here.