Refiberd AI Hyperspectral Textile Material Detection

Refiberd, Inc. is a US company led by a women-founded team of engineers, Sarika Bajaj (CEO, formerly working on fashion-tech at Intel) and Tushita Gupta (CTO, an AI scientist formerly at Atomwise). It addresses textile waste and material fraud: recyclers lose yield from poorly sorted feedstock, brands lose value when materials are switched in production, and chemical lab tests for fiber content are slow and destructive.

Its system places a garment or textile under a hyperspectral camera and line light source; as the item moves past, the camera captures lines of spectral data that a computer stitches into a hyperspectral cube in which each pixel is a spectrum. A machine learning model trained on a proprietary dataset of thousands of custom textile samples predicts the material composition, including blended and layered fibers, and flags contaminants. The technology targets textile recyclers, sorters, resellers and brands that need fast, non-destructive fiber identification.

• Hyperspectral imaging: a hyperspectral camera and line light source scan textiles as they move past.
• Hyperspectral cube: scanned lines are stitched into a cube where each pixel holds a full spectrum.
• AI composition model: machine learning trained on thousands of custom textile samples predicts fiber content.
• Blend and layer detection: identifies all fibers present, including blended and layered materials.
• Trace contaminant detection: flags spandex, nylon and acrylic below 2% composition.
• Non-destructive testing: contact-free analysis leaves garments intact for resale or recycling.

Textile recycling plants in Vietnam could screen incoming cutting waste to separate elastane-containing material before mechanical or chemical recycling. Garment factories and waste collectors could grade fabric scraps by composition to sell higher-value sorted feedstock. Brands and quality labs could spot-check fiber content of fabrics supplied by mills.

Refiberd states contact-free detection at millisecond speed per garment, versus slow and destructive chemical analysis. It can detect trace spandex, nylon and acrylic below 2% composition, the kind of trace contaminant that affects recycling feedstock quality. Accurate composition data helps recyclers improve feedstock purity and yield and helps brands verify materials delivered by suppliers.

Fiber-to-fiber recyclers need feedstock sorted precisely by composition, and EU textile collection and producer responsibility rules are increasing volumes that must be sorted. Vietnam's large garment sector generates substantial cutting waste and is beginning to attract textile recycling plants, creating demand for automated composition identification.

Legal entity Refiberd, Inc., with terms governed by US law. Co-founders Sarika Bajaj (CEO) and Tushita Gupta (CTO) both studied at Carnegie Mellon. Winner of the 2025 CFDA x eBay Circular Fashion Fund. Founding year and headquarters city are not stated on the website.