Personalized AI Recommendations for Website

I want to enrich my Arabic-coffee e-commerce site with an AI-powered recommendation engine that suggests products each visitor is most likely to love. The core feature is product recommendations—specifically personalized suggestions generated from a user’s browsing and purchase behaviour rather than generic “bought together” lists. These tailored offers should appear on individual product pages and on the homepage carousel so shoppers instantly see items that fit their taste the moment they land or browse.

The site is already live on WordPress with WooCommerce, so the solution needs to plug into that stack smoothly via plugin, API, or lightweight custom microservice. Real-time performance is essential; latency above a second will hurt conversions. Because the brand story and UI are in Arabic, all visible text from the AI module must render in Arabic while remaining UTF-8 friendly.

Key deliverables:
• Data hookup that securely streams click, view, and purchase events to the model
• Recommendation algorithm (ready-made SDKs like TensorFlow Recommenders or a SaaS such as Amazon Personalize are fine, as long as it can be tuned)
• Front-end widgets for the product pages and homepage matching current site design
• Admin controls to adjust model parameters and track performance (CTR, AOV)
• Clear setup and hand-off documentation

Acceptance criteria:
1. Recommendations refresh based on new user actions within the same session.
2. Widgets load in <1 s on a 4G connection.
3. All Arabic copy displays correctly without encoding issues.
4. A/B toggle available so I can measure lift against the current static upsells.

Please mention any prior projects where you implemented personalized recommendations, the tech stack you prefer for the model, and an estimated timeline for deployment and testing.