AI robots and optical sorters for materials recovery facilities.
Photo credit: Recycleye
Recycleye builds retrofittable AI vision for waste sorting: six-axis robots and optical/air-jet systems that identify objects in dry mixed recyclables, WEEE and construction debris. The company was founded in 2019 in London by CEO Victor Dewulf and CTO Peter Hedley, who started with a camera, a treadmill and dumpster-dived trash while Dewulf pursued machine-vision research after leaving investment banking. Academic roots include Imperial College London and Delft.
A UK partnership with FANUC supplies industrial arms. In April 2026, San Diego-based CP Group acquired a majority stake after more than 23 joint installations of Recycleye vision with CPG optical sorters in the US and Europe. Recycleye’s London engineering team remains in place, combining AI classification with near-infrared ejection and robotics so MRFs can lift purity without a full plant rebuild.
Recycling fails when the cost of sorting exceeds the value of the commodity. Recycleye attacks that gap with computer vision that sees hundreds of material classes and then either picks (robots) or ejects (QuantiSort / Vivid-AI optical). Higher recovery of PET and other fractions, plus less hazardous manual picking, is the impact pathway.
The CP Group majority acquisition is meant to industrialise that vision inside full MRF platforms rather than as a standalone gadget, while keeping Recycleye as a distinct AI division for Europe and North America.
* Founded 2019; HQ 11-15 Gibbins Road, London E15 2HU
* About $24M venture funding before the 2026 CP Group deal, including $17M Series A (Feb 2023, DCVC) plus earlier seed and ~$2.6M UK/EU innovation grants
* Series A-era deployments already spanned England, Ireland, Scotland, Germany, Australia, the US and France
* 23+ live installations with CP Group optical sorters by the April 2026 acquisition announcement
* ~45-person team including engineers from Cambridge and Imperial (CP Group announcement)
* Vision at 60 frames/sec; up to ~33,000 picks per robot per 10-hour shift (2023 Series A announcement)
* Founders awarded joint first place in the EPO Young Inventors Prize
Retrofittable robots
FANUC-based six-axis cells add picking to existing belts without a greenfield plant.
AI optical sorting
Object-level models drive air-jet and optical units across DMR, WEEE and C&D streams.
Plant-builder distribution
CP Group’s majority stake pairs Recycleye vision with North American MRF equipment platforms.
Research pedigree
WasteNet research and Imperial/Delft collaboration trained models on real dumpster-sourced waste.
Focus areas
Locations
