Project Hub
Start from the project, not the tool
The dashboard made Bloom Hub feel like a workspace: create a project, reopen it, and keep design intelligence attached to that project instead of scattered across one-off tools.
A private architecture workspace for design and project intelligence
I built Bloom Hub as a private design-and-information hub for Bloom Architects: a place to start projects, test AI design tools, run zoning and site analysis, manage documents, and keep project intelligence in one workspace.
Private Hub
Built as a Bloom Architects workspace with project context at the center.
Design Sandbox
AI ideation, program massing, and sketch-to-3D layout generation lived inside one project context.
Information Layer
Zoning, ZoningPal reports, site finding, documents, and project chat sat beside the design tools.
Full Stack
React, Three.js, FastAPI, Supabase, PostGIS, CadQuery, Gemini Vision, GLB and IFC-oriented output.
Product Idea
The hub mattered because architecture work is fragmented.
One tool can generate an image. Another can show a map. Another can store a PDF. Bloom Hub tried to make those separate actions feel like one project workspace.
Frontend
A real workspace shell
React, Vite, and React Three Fiber powered a project dashboard, authenticated routes, feature modules, layer controls, selectable volumes, imported models, camera state, and 3D interaction.
Engines
Design tools behind the interface
The workspace combined concept generation, layout processing, zoning, site finding, assets, and concierge features with solar/shadow and wind-rose overlays for design context.
Geometry
Sketch-to-model pipeline
The layout pipeline moved from wall tracing and space segmentation into scale detection, room labeling, parametric geometry, GLB preview, and BIM/IFC-oriented output.
Data
Project and city intelligence
Supabase handled auth, project state, saved iterations, camera state, uploaded assets, reports, and conversation records. PostGIS supported zoning and parcel-scale site-finding queries.
Workspace Modules
Design tools and information tools in the same place
Sketch-to-3D was one strong module inside a larger architecture workspace for design exploration, feasibility, project context, and AI assistance.
Design Studio
A sandbox for early design moves
The design studio grouped ideation, concept massing, and layout generation. That matters because early design is not one linear feature. It is a loop between mood, mass, geometry, and constraints.
Zoning
ZoningPal inside the project context
The zoning module let the workspace query property information and generate ZoningPal reports without leaving the Bloom Hub project environment.
Site Finder
Search the city as design input
The site finder turned parcel dimensions and zoning filters into a design-research tool. It let Bloom look across Toronto before committing to a specific site.
Design Workflow
From early design exploration to structured model output
The design side had to feel visual and fast, but the technical stack underneath was about structure: project state, editable 3D objects, semantic layout data, and exportable geometry.
Sketch-To-3D Architecture
The hard part was turning ambiguity into structure.
A floor-plan sketch is messy: lines are uncertain, dimensions can be partial, and meaning is not explicit. Bloom Hub treated that as a pipeline problem, not a magic button.
Why It Matters
This sits in the same problem category as mature design-automation tools, but the claim here is narrower: a functional solo-built core architecture for sketch input, geometry extraction, semantic interpretation, model generation, and project persistence.
01 / Ingest
Make the sketch part of the project
The upload flow creates a layout record, stores the source sketch, and starts a background processing job instead of treating the image as a loose file.
02 / Extract
Separate signal from noise
The pipeline traces walls, segments spaces, vectorizes geometry, detects rooms, reads dimensions, and aligns space polygons back to wall structure.
03 / Interpret
Add architectural meaning
AI room labeling and wall classification turn raw geometry into semantic layout data: spaces, wall types, openings, scale, and usable project metadata.
04 / Generate
Build model output
Parametric walls, floors, ceilings, spaces, doors, and windows are assembled into a GLB preview with a BIM/IFC-oriented path behind it.
05 / Persist
Return it to the workspace
The model URL, floor-plan JSON, walls, spaces, thumbnail, status, and stats are saved back into the project context so the output can be reused.
Zoning + Site Intelligence
The information side was as important as the design side.
Bloom Hub connected the ZoningPal report flow and a Toronto parcel finder into the same workspace. That makes the product story stronger: design decisions and feasibility information lived beside each other.
Scope
Built for one practice, but broad in scope.
Bloom Hub was built for a specific architecture practice, but the system itself was broad: design tools, site intelligence, zoning context, documents, and project-aware assistance.
Private Workspace
A focused workspace for one architecture practice
Bloom Hub was built around the needs of Bloom Architects: early design exploration, site intelligence, zoning context, project documents, and AI assistance in one place.
System Architecture
A private AEC product that connected design tools and information tools
The value is the product architecture: project context, AI design workflows, geometry generation, zoning intelligence, site search, storage, and chat in one workspace.
Signal
Project context can hold design tools and information tools together.
Bloom Hub shows both halves of the work: understanding what architects actually need, then building the interface, data model, geometry pipeline, and services that let those workflows live together in one place.