Faddi Idris

At Ingka Digital (IKEA), I worked across recommendation products, data, and engineering. For Content Recommendations, I found that visitors without browsing history were seeing random items, then built per-market popularity lists that increased clicks through to products by about 18% in an A/B test.

In Product Recommendations, I built a static cart-page baseline that raised the number of visitors adding from the panel by about 20%, then designed and built a Thompson-sampling bandit that improved on that baseline by about 20%. Both results were statistically significant, and the bandit rolled out to about 30 markets.

I co-founded Decision Orchestration, a shared recommendations platform, and contributed its multi-region serving layer and trace-ingest path. I also proposed and built a general-purpose bandit so frontend teams could bring their own candidates and receive rankings that adapt online.

Earlier, I took IKEA's text-to-image asset search from an experiment to a production tool that remains in use. At Strusoft, I turned structural-engineering calculations into desktop and web software, often as a product's only developer.