AI-powered product libraries for construction using semantic data and vision models
Depixen builds autonomous product libraries for the construction ecosystem using PyTorch, TensorFlow, and vision models (YOLO, Detectron2, Vision Transformer) paired with semantic technologies (SPARQL, knowledge graphs). The tech stack reveals a dual focus: computer vision for product perception and linked-data systems for ecosystem coordination. Active hiring is senior-heavy (7 of 9 roles) across engineering and data, concentrated in Turkey, while projects span ML-driven classification, payment infrastructure, and semantic search — suggesting both technical depth and operational complexity as they scale.
Depixen is a London-based AI and construction-software company founded in 2008. The company designs autonomous product libraries that aggregate data from the construction supply chain — over 200 manufacturers and 100+ architects, designers, and consultants — using semantic technologies aligned with W3C standards. Their platform combines computer vision for product recognition with knowledge graphs and ontologies to create a unified, machine-readable view of construction products and their properties. Revenue streams include payment and wallet services alongside data-driven licensing models. The 11–50 person team is engineering-led, with active hiring in senior technical roles.
PyTorch, TensorFlow, YOLO, Detectron2, Vision Transformer, MLflow, Weights & Biases, and SPARQL for semantic reasoning. The company is adopting CLIP for vision-language alignment.
Autonomous product libraries, next-generation perception systems, semantic data frameworks (ontologies, knowledge graphs), ML-driven product classification, payment infrastructure, and semantic search capabilities for the construction ecosystem.
Other companies in the same industry, closest in size