Building an AI learning assistant into a digital education platform.

The challenge

A learning assistant has to do more than respond to general questions. It needs to connect the learner's request with relevant educational material and deliver that interaction within the platform where learning is already taking place.

For the education provider, this meant bringing conversational AI into an existing product architecture. Content access, retrieval, model interaction and application services needed to work together without turning the assistant into an isolated demonstration. The platform also needed a way to evolve AI behaviour as models, prompts and content changed.

The solution

The work architected retrieval-augmented AI capabilities for a conversational learning assistant. GPT-based model interaction was connected to the platform's content and service layer, enabling learning assistance grounded in the educational material available to the system.

The architecture combined TypeScript and Go microservices with GraphQL Federation, exposing related backend capabilities through a cohesive API. Asynchronous processing supported background work such as content preparation and indexing, while AWS deployment infrastructure supported the application and its operational services.

The AI foundation included hybrid semantic and keyword retrieval, using embeddings and search services to find relevant material. This connected the tutoring experience to a wider content-discovery architecture rather than depending on the model's general knowledge alone. GPT-based tutoring was integrated with the wider AWS-hosted application architecture. The project was subsequently handed over to the education provider's internal AI management team.