Sewise Fashion Labs: Made-to-Measure Sewing Patterns

A 2026 hackathon project that turns a garment sketch and five body measurements into a made-to-measure pattern, calculated fabric lengths for each roll width and a shopping list, with tools for reusing old clothes.

1,390
Backend tests passing
118
Commits by Kurniadi in 40 pull requests
9
Feature specs, 754 planned tasks

Role: Built for a 2026 hackathon (Deep Domain AI direction) and planned for a four-person team; Kurniadi wrote all the code (118 commits, 40 pull requests) with GitHub Copilot and GitHub Spec Kit
Period: Oct 2026
Project type: Featured
Room in the 3D HQ: L3:sewise-fashion-labs

Results

1,390
Backend tests passing
pytest without the live-model tests, run on a fresh clone, 16 Oct 2026.
118
Commits by Kurniadi in 40 pull requests
Repository history, 9 to 10 Oct 2026.
9
Feature specs, 754 planned tasks
GitHub Spec Kit overview in the repository.

Problem

Home sewers, small tailors and fashion students in Indonesia need pattern-making skills to turn a design into a pattern that fits. Fabric is bought from generic charts that ignore the roll width, notions get forgotten, and old clothes are thrown away even when they could be reused.

Contribution

I put pattern dimensions in deterministic, tested code. A Python drafting engine calculates the bodice, sleeves, peplum and skirt from five measurements. A fit report compares each pair of adjoining seams with its target in millimetres: checks of calculated geometry, not physical garment-fit validation.
A nesting engine lays the pieces out on 110, 115 and 150 cm fabric with grainline rules. It falls back to a simpler layout heuristic when the main engine is unavailable, then uses the calculated fabric lengths in a shopping list with thread, zipper and needle.
AI only reads intent and orchestrates: Azure OpenAI reads a sketch into a constrained design spec, and an assistant on Microsoft Agent Framework calls the deterministic services as tools. Without the model, the samples and the manual spec form keep the whole flow working.
I froze the API contracts first (Pydantic models, OpenAPI and generated TypeScript types), then built nine feature specs with GitHub Copilot and GitHub Spec Kit on Cosmos DB, Blob Storage and Azure Container Apps scripts that need only Contributor access.