Daniel Howells

Materia

2024– · React, Next.js, TypeScript, tRPC, Drizzle, pgvector

Product search platform with vector image search, AI recommendations, and a comprehensive design system

Architected the frontend platform for Materia, a materials marketplace serving architects and designers. Built the product search system — vector-based image search across millions of material samples, a colour search engine using quantised HSL and OKLab models, and AI-powered product recommendations.

Led the frontend architecture: a 33-package monorepo with five apps, end-to-end type safety from Drizzle ORM through tRPC to React Query, and a design system built on Base UI primitives. The database schema alone is over a thousand lines — products with 165 JSONB attributes, HNSW vector indexes, hierarchical categories via ltree, and a dual colour model for perceptual matching.

Rapid prototyping is a core part of the role. I've prototyped and shipped a mobile NFC tagging experience, a material comparison tool, a palette-based search interface, and an AI-assisted product categorisation pipeline. Most features go from idea to working prototype within a day.