SilkRoute AI

• 

  SilkRoute Dashboard

• 

  REciept

Inspiration

Over \$500 billion worth of high-quality textile offcuts, selvages, and roll ends are sent to landfills or incinerators annually. In luxury apparel and couture manufacturing, up to 25% to 40% of raw silk, jacquard, and brocade fabrics end up as pre-consumer cutting waste.

Small artisan clusters and circular accessory makers struggle to find consistent, high-grade scrap lots, while apparel factories lack automated tools to grade, price, and distribute odd-shaped offcuts efficiently. SilkRoute AI was born to bridge this gap: an automated B2B circular intelligence and inventory recovery engine that transforms fabric waste into verified revenue.

What It Does

SilkRoute AI automates the end-to-end circular textile supply chain through three core pillars:

1. Multimodal Material & Spatial Diagnostics: Users upload or capture high-resolution photos of fabric scraps. Our vision model analyzes fiber composition, thread count, GSM density ($\text{g/m}^2$), colorimetric spectrum (HEX/RGB), and calculates exact usable polygonal surface area: $$\text{Yield} = \frac{A_{\text{usable}}}{A_{\text{total}}} \times 100\%$$
2. Automated B2B Scrap Lot Valuation: Automatically calculates fair wholesale market pricing using thread density, fiber purity, and local market spot rates, matching lots with regional micro-manufacturing buyers.
3. EU Digital Product Passport (DPP) & ESG Telemetry: Computes ISO 14044-compliant Life Cycle Assessment (LCA) data (liters of water preserved, $\text{kg CO}_2\text{e}$ mitigated, and landfill mass diversion) and generates audit-ready QR passports.

How We Built It

• Frontend Architecture: Built with React 18, TypeScript, and Tailwind CSS, featuring dark-slate enterprise layouts, interactive colorimetric pickers, dynamic modular view filters, and live telemetry data tables.
• Backend & AI Diagnostic Pipeline: Express.js and Node.js integrated with Google Gemini 2.5 Flash for rapid multimodal vision inference and structured JSON material schema extraction.
• Standards & Calculations Engine: Embedded mathematical models for thread density, GSM estimation, carbon offset factors ($E_{\text{CO}2} = m{\text{diverted}} \times F_{\text{virgin}}$), and water savings footprints.
• Deployment & Cloud Infrastructure: Packaged with Vite and esbuild into a production CommonJS bundle deployed on Google Cloud Run.

Challenges We Faced

• Polygonal Fabric Geometry Estimation: Estimating true cutting yield from irregular scrap shapes required carefully engineered vision prompts to distinguish frayed selvages from clean usable fabric.
• Multi-Currency & Regional Valuation: Balancing real-time INR (₹) domestic artisan cluster pricing with USD (\$) export market valuations.
• Enterprise-Grade CommonJS Bundling: Fine-tuning the backend build pipeline with esbuild and Vite middleware to guarantee zero-latency container cold-starts on Google Cloud Run.

What We Learned

• High-precision multimodal models can accurately identify complex micro-structures in luxury fabrics (such as Banarasi Zari, Mulberry Silk, and Ikat patterns) from standard camera captures.
• Providing standardized EU Digital Product Passport (DPP) data creates immediate commercial value for B2B brands preparing for upcoming EU circular economy regulations.

What's Next for SilkRoute AI

• Direct ERP Integration: Connecting with SAP and Lectra/Gerber automated cutting tables to log offcuts automatically during fabric cutting.
• Automated Physical QR Tag Generation: Generating thermal printable QR sticker manifests for warehouse scrap bin routing.
• Expanded Hyper-Local Artisan Networks: Onboarding additional verified weaving and upcycling clusters across South Asia, Europe, and Latin America.

Built With

• canvas-confetti
• circular-economy
• cleantech
• computer-vision
• digital
• esg-analytics
• express.js
• google-cloud-run
• google-gemini-api
• json-schema
• life-cycle-assessment-(lca)
• lucide-icons
• node.js
• passport
• product
• react
• rest-apis
• supply-chain-ai
• sustainability
• tailwind-css
• typescript
• vite