I build AI features that run in live stores, not demos. I am an AWS Certified AI Practitioner, and the image search, support chatbot and catalog tools described below are in production.
Useful AI in ecommerce mostly comes down to your own data: products, photos, order history and support answers. The work is getting that data into a form a model can use, choosing a model that fits the budget, and checking that the answers are right before customers see them.
What I can help with
- Image and semantic search. Shoppers upload a photo or describe what they want, and vector embeddings find similar products across the catalog.
- Support chatbots. RAG chatbots that answer from your product data, FAQs and order history. On one live store this reduced the number of support tickets.
- Catalog intelligence. Finding duplicate and near-duplicate products from their photos with image embeddings, and tidying large catalogs.
- AI in the admin. Tools such as MageAsk, which answers questions about store data in plain English.
- Workflow automation. AI steps added to processes you already run, such as drafting product content or sorting incoming enquiries.
Related work
- MageAsk AI Powered Admin Assistant, my Magento extension built on OpenAI.
- AI Model Comparison: closed and open models for chat, images, video and speech, side by side.
Stack
AWS Bedrock, SageMaker, OpenAI, vector databases (pgvector, Qdrant), RAG, Python
Questions
Which AI models do you use?
Whichever fits the job and the budget: models on AWS Bedrock, OpenAI, or open-weight models you host yourself. The AI Model Comparison tool shows the options and their prices.
Will our data be used to train someone else’s model?
That depends on the provider and plan. I choose providers and settings that keep your data out of training, or host an open model on your own servers when data must not leave them.