expertise / architecture

Architecture training data. Plans that would pass a plan check.

A generated layout can look finished and still have a stair that lands in a wall. eeko commissions practicing architects and architectural designers to label plans, classify building models, write spatial reasoning items, and critique generated designs, with rights recorded on every drawing and a second architect on every item.

Two architects on every item Rights recorded per drawing Pilot batch before volume Reply inside one business day
Hero image · 840 × 1050
reviewers per item
2arch
the annotator, then a second architect
first delivery
1pilot
a paid pilot batch before any volume order
delivery formats
6fmt
JSON Lines, CSV, Parquet, WebDataset, COCO, Croissant
reply time
1day
a data lead who has read the brief

What is architecture training data?

Architecture training data is design material with the architect's understanding attached: a floor plan where every room, wall, door, window, and circulation path carries a label; a BIM model where each element is classified and its properties recorded; a rendering described by style, material, and spatial quality; a generated layout with a written critique explaining why the egress fails and the bedroom has no daylight. The understanding is what the model is buying. The drawing on its own is just pixels or geometry.

Models trained on scraped plan images learn what plans look like and nothing about how buildings work. The results are familiar to anyone who has evaluated them: doors that swing into each other, a corridor that dead-ends at a bathroom, a two-bedroom unit with one window, rooms drawn at a scale no furniture would fit. Each is convincing at a glance and useless to anyone who has sat through a plan check.

eeko's data practice puts architects in the loop at the point where it counts. They label to a written schema, they note which jurisdiction's assumptions apply so a model does not generalize a code it never read, and a second architect checks every item before delivery. It is one field within our AI training data practice, and the natural partner to our 3D and CAD datasets and image datasets.

Annotated plan image · 2100 × 900
A residential plan after annotation: rooms by use, walls by type, openings by swing, circulation as a path, dimensions checked against the scale bar, every tag attributed to the architect who placed it.

Architectural data for AI, from plan to critique.

Six lines of work, each scoped by building type, jurisdiction, and format, and proven on a pilot batch. Combine them or order one layer on drawings you already hold.

01 / plansCORE
Annotated floor plans
Plans labeled for rooms and their use, walls by type, doors and windows with swing and size, circulation paths, dimensions, and level, following a written schema that two annotators apply the same way.
02 / setsCORE
Drawing sets
Plans, sections, and elevations for the same building, licensed from the practice that drew them or drawn new for your request, with labels that stay consistent from sheet to sheet.
03 / bimCORE
BIM and 3D models
Building models with elements classified, properties recorded, and relationships preserved, fingerprinted by shape and delivered in native and open formats.
04 / critiqueEVAL
Critique of generated designs
Architects review AI-generated plans and massing for egress, accessibility, daylight, structure, and buildability, and write the critique a plan checker would. Scored against a rubric, with the jurisdiction noted.
05 / reasoningEVAL
Spatial reasoning items
Questions about layouts, adjacency, scale, orientation, and circulation, with worked answers, for evaluating and fine-tuning multimodal models. Written by architects, answered by architects, checked by a third.
06 / imagesCORE
Rendering and photography labels
Architectural photographs and renderings described by style, period, material, element, and spatial quality, with consent from the photographer or renderer attached to each image.

What a generated plan gets wrong and an architect sees first.

Every one of these passes an image-similarity metric. None would pass a plan check. Our critique rubric is built around this list.

Critique image · 1200 × 900
A generated layout under review: egress path traced, door swings checked, daylight noted, and the jurisdiction's assumptions written in the margin.

The failures the critique rubric is built on

  • No way out — a stair that arrives nowhere, a corridor that dead-ends, travel distances no code would accept.
  • Openings that collide — doors swinging into each other, a window into a party wall, a door where a column has to be.
  • Rooms without light — a habitable room with no window, or a window that faces a lightwell the model invented.
  • Scale that does not work — a kitchen nobody could stand in, a bedroom that fits a bed and nothing else, corridors wider than the rooms.
  • Circulation that ignores use — a bathroom entered from the dining room, a bedroom reached through another bedroom.
  • Code applied from nowhere — egress and accessibility assumptions that belong to a jurisdiction the brief never mentioned.

Paid remote work for architects, between projects.

The eeko expert network pays architects and designers to annotate, classify, critique, and review. Rate, scope, and deadline are on every job. It is data work, not design services: nothing is stamped.

who we look for
Professional training, practice experience

You do not need every item below. A professional degree plus practice experience is enough to start receiving matching jobs.

  • Degree BArch, MArch, or an equivalent professional qualification
  • Registration Licensed or registered, for critique and code-related tasks
  • Practice Years in residential, commercial, or institutional work
  • Tools Fluent in BIM and CAD authoring software
  • Specialism Accessibility, housing, healthcare, interiors, or urban design
what the jobs involve
Your judgment, applied to a schema

Written schema or rubric on every job, a second architect on every item, and approved work credited to your account automatically.

  • Annotate Rooms, walls, openings, and circulation paths on plans and sections
  • Critique Generated layouts for egress, access, daylight, and buildability
  • Write Spatial reasoning questions with correct, reasoned answers
  • Review Another architect's labels and critiques before delivery

Rights, fingerprints, and review on every drawing.

Drawings and models are creative work with an author. A model trained on them must show where they came from and who agreed.

Licensed or drawn new

Every plan, model, and rendering is licensed from the practice or designer who holds the rights, or produced by an architect in the network under terms stated in the job. The licence and consent are attached to the individual file, not the batch.

Fingerprinted and screened

Images receive a perceptual hash, models a 3D shape fingerprint, and both a cryptographic hash and an independent timestamp on intake. Embedded metadata is read to confirm human origin, and duplicates are rejected before an architect sees the file.

Attributed and receipted

Every label and critique records who made it and who confirmed it, inside a hash-chained audit trail. Delivery closes with a signed receipt naming each item by identifier and hash. The method is described in full on our verified human data page.

From a building type to a signed delivery.

Fixed price per accepted item, agreed on the pilot and held for production. Volume starts when your team has reviewed real output against the schema.

STEP 01
Scope
Building types, jurisdictions, label layers, source of drawings, formats, and licence terms. A data lead replies inside one business day with questions and a staffing view.
STEP 02
Schema and pilot
We write the labeling schema or critique rubric with you, then deliver a paid pilot batch so you can check label depth, consistency, and format.
STEP 03
Production
Verified architects work in batches. A second architect reviews every item, disagreements are adjudicated and logged, and agreement is reported per delivery.
STEP 04
Delivery
COCO or WebDataset for images, native and IFC for models, JSON Lines or Parquet for labels, Croissant metadata, and a signed receipt. Standing orders add building types.

Architecture training data questions.

What is architecture training data?

Architecture training data is labeled design material a spatial or generative model learns from and is tested against: floor plans with every room, wall, opening, and circulation path tagged, BIM models with elements classified, renderings described by element and material, and generated layouts critiqued by an architect for whether a person could actually get out of the building. The architect behind the label is what separates it from scraped plan images.

How much does architecture training data cost?

We price per accepted item, and the rate depends on the task and the depth of the label. Tagging rooms on a residential plan costs less than classifying a full BIM model or writing a reasoned critique of a generated hospital layout, and work that needs a registered architect costs more. After scoping we quote a fixed price for a pilot batch, and the same unit price carries into production. Rejected items are not billed.

What kinds of architectural data can you supply?

Annotated floor plans, full drawing sets, BIM and 3D models with classified elements, written critiques of AI-generated designs, spatial reasoning questions with worked answers, and labeled renderings and photographs. Each request is scoped by building type, jurisdiction, scale, and format, then staffed with architects verified in that kind of work. The pilot batch confirms the schema before any volume is agreed.

Who holds the rights to the drawings and models you deliver?

Either the drawings are licensed from the practice or designer who holds the rights, with the licence and consent attached to each file, or they are drawn new by architects in the network for your request under terms stated in the job. Every image gets a perceptual hash and every model a shape fingerprint on intake, embedded metadata is read to confirm human origin, and nothing is scraped from portfolios or planning portals.

Is a license required to join, and how are architects paid?

Not for every job. Plan annotation, element classification, and rendering labels are open to architectural designers who hold a professional degree and have practice experience, while critique and code-dependent tasks generally call for registration. The work is data for AI, not design services, so nothing is stamped. Every job lists its rate, scope, and deadline before you accept it, and approved batches are credited to your account with payouts after identity and tax verification.

Can we license the plans and models exclusively?

Yes. Commissioned drawings, models, labels, and critiques can be licensed to one buyer, with scope, term, and exclusivity written into the agreement before work starts. Images ship as COCO or WebDataset, models in native formats with Parquet or JSON Lines metadata, and every delivery includes Croissant metadata and a signed receipt naming each item by identifier and hash.

Show us the plan your model is proudest of.

Send the building types, label layers, and volume. A data lead replies inside one business day with a staffing view, a schema outline, and a pilot price. Architects: the application takes ten minutes. No deck. A ranked list.

Experienced within

  • SOC 2 Type II
  • ISO 27001
  • HIPAA
  • HITRUST r2
  • ISO 27701
  • PCI DSS
  • NIST CSF 2.0
  • CMMC
  • FedRAMP
  • GDPR

Markets served.

As an enterprise AI agency, eeko systems delivers production AI systems remote-first across the United States and internationally — including these markets:

New York City, New York (NY)

Los Angeles, California (CA)

Chicago, Illinois (IL)

Houston, Texas (TX)

Phoenix, Arizona (AZ)

Philadelphia, Pennsylvania (PA)

San Antonio, Texas (TX)

San Diego, California (CA)

Dallas, Texas (TX)

San Jose, California (CA)

Austin, Texas (TX)

Jacksonville, Florida (FL)

Fort Worth, Texas (TX)

Columbus, Ohio (OH)

Charlotte, North Carolina (NC)

Indianapolis, Indiana (IN)

San Francisco, California (CA)

Seattle, Washington (WA)

Denver, Colorado (CO)

Washington, District of Columbia (DC)

Boston, Massachusetts (MA)

El Paso, Texas (TX)

Nashville, Tennessee (TN)

Detroit, Michigan (MI)

Oklahoma City, Oklahoma (OK)

Portland, Oregon (OR)

Las Vegas, Nevada (NV)

Memphis, Tennessee (TN)

Louisville, Kentucky (KY)

Baltimore, Maryland (MD)

Milwaukee, Wisconsin (WI)

Albuquerque, New Mexico (NM)

Tucson, Arizona (AZ)

Fresno, California (CA)

Sacramento, California (CA)

Kansas City, Missouri (MO)

Atlanta, Georgia (GA)

Miami, Florida (FL)

Colorado Springs, Colorado (CO)

Raleigh, North Carolina (NC)

Omaha, Nebraska (NE)

Long Beach, California (CA)

Virginia Beach, Virginia (VA)

Oakland, California (CA)

Minneapolis, Minnesota (MN)

Tulsa, Oklahoma (OK)

Arlington, Texas (TX)

New Orleans, Louisiana (LA)

Wichita, Kansas (KS)

Cleveland, Ohio (OH)

Tampa, Florida (FL)

Bakersfield, California (CA)

Aurora, Colorado (CO)

Honolulu, Hawaii (HI)

Anaheim, California (CA)

Santa Ana, California (CA)

Corpus Christi, Texas (TX)

Riverside, California (CA)

Lexington, Kentucky (KY)

St. Louis, Missouri (MO)

Stockton, California (CA)

Pittsburgh, Pennsylvania (PA)

Saint Paul, Minnesota (MN)

Cincinnati, Ohio (OH)

Greensboro, North Carolina (NC)

Anchorage, Alaska (AK)

Plano, Texas (TX)

Lincoln, Nebraska (NE)

Orlando, Florida (FL)

Irvine, California (CA)

Newark, New Jersey (NJ)

Toledo, Ohio (OH)

Durham, North Carolina (NC)

Chula Vista, California (CA)

Fort Wayne, Indiana (IN)

Jersey City, New Jersey (NJ)

St. Petersburg, Florida (FL)

Laredo, Texas (TX)

Madison, Wisconsin (WI)

Chandler, Arizona (AZ)

Buffalo, New York (NY)

Lubbock, Texas (TX)

Scottsdale, Arizona (AZ)

Reno, Nevada (NV)

Glendale, Arizona (AZ)

Gilbert, Arizona (AZ)

Winston-Salem, North Carolina (NC)

North Las Vegas, Nevada (NV)

Norfolk, Virginia (VA)

Chesapeake, Virginia (VA)

Fremont, California (CA)

Garland, Texas (TX)

Richmond, Virginia (VA)

Baton Rouge, Louisiana (LA)

Boise, Idaho (ID)

San Bernardino, California (CA)

Spokane, Washington (WA)

Des Moines, Iowa (IA)

Modesto, California (CA)

Birmingham, Alabama (AL)

Tacoma, Washington (WA)

Fontana, California (CA)

Oxnard, California (CA)

Fayetteville, North Carolina (NC)

Huntsville, Alabama (AL)

Moreno Valley, California (CA)

Rochester, New York (NY)

Glendale, California (CA)

Yonkers, New York (NY)

Augusta, Georgia (GA)

Amarillo, Texas (TX)

Little Rock, Arkansas (AR)

Akron, Ohio (OH)

Shreveport, Louisiana (LA)

Grand Rapids, Michigan (MI)

Mobile, Alabama (AL)

Salt Lake City, Utah (UT)

Huntsville, Texas (TX)

Tallahassee, Florida (FL)

Overland Park, Kansas (KS)

Knoxville, Tennessee (TN)

Worcester, Massachusetts (MA)

Brownsville, Texas (TX)

New Port Richey, Florida (FL)

Jackson, Mississippi (MS)

Providence, Rhode Island (RI)

Fort Lauderdale, Florida (FL)

Sioux Falls, South Dakota (SD)

Tempe, Arizona (AZ)

Cape Coral, Florida (FL)

Springfield, Missouri (MO)

Pembroke Pines, Florida (FL)

Eugene, Oregon (OR)

Peoria, Arizona (AZ)

Corona, California (CA)

Lancaster, California (CA)

Rockford, Illinois (IL)

Salinas, California (CA)

Palmdale, California (CA)

Springfield, Massachusetts (MA)

Charleston, South Carolina (SC)

Duluth, Minnesota (MN)

London, England (ENG)

Dublin, Ireland (IRE)