AI in e-commerce: search, product discovery and platform design
How AI can support search, ranking and product decisions, and the catalogue, integration and evaluation work needed to make those features useful.
Case studies, articles, and lessons from the engineers and teams at Outmatic.
How AI can support search, ranking and product decisions, and the catalogue, integration and evaluation work needed to make those features useful.
Vespa.NET is our open-source .NET client for Vespa, covering schemas, feeding and typed queries. This post covers version 3.0.0 and its breaking changes for users upgrading from 2.0.0.
We have released VoyageAI.NET 1.0.0, our MIT-licensed .NET client for Voyage AI embeddings and reranking, with source-generated JSON metadata, batching and standard HTTP resilience.
We have built and open-sourced TurboQuant for .NET. Here is what is in the package, where it can help, and why we chose native bfloat16 for a client using Vespa.
A stock integration replaced twice-daily updates for an Italian footwear-materials supplier. It reduced orders cancelled or amended for stock shortages by 95% and cost roughly 70% less than the proposed e-commerce rebuild.
For Experia Group’s client, an Italian knitwear retailer with dozens of stores, we built a promotions engine that lets business teams define campaigns and run their rules locally at the checkout.
We built a loyalty platform for Experia Group’s client, an Italian jewellery brand with more than 50 stores. It shares points and discount tiers with e-commerce and returns discounts within the POS system’s 150 ms window.
We modernised Bunge’s existing Permit to Work application, improved its main workflows and added mobile signatures while retaining the system used across its plants.
We built a Permit to Work platform for Tampieri’s industrial sites, bringing approvals, role-based access and signatures on phones and tablets into one system.
We connected Clictravel with Alpitour, Eden, Ideal and Nicolaus, automating daily package updates and adding live availability and pricing queries at search time.
We extended Davidoff’s existing platform for the United States, Switzerland and Germany, covering performance, checkout, content and semantic search. Results included +45% average page load speed and –60% time spent managing content.
We built a shared test automation framework for Credemtel’s applications. The project reduced manual testing time by 70% and increased functional coverage by 50%.
We worked with BKN301 to build 301pay, connecting payment networks, banking systems and digital channels, and helped its internal team take ownership of the platform.