Internship · Sofrecom Tunisia2025RAG Low-code UI Generator
Can a natural-language prompt produce a usable low-code interface?
During my summer internship as an AI intern at Sofrecom Tunisia, I built a RAG (Retrieval-Augmented Generation) system that automatically generates user interfaces in GrapesJS from natural-language prompts, to speed up low-code application development. An innovative project bringing generative AI and low-code together.
- My role
- AI intern: designed and built the RAG system and its prompts
- Duration
- 2 months (Jul – Aug 2025)
- Platforms
- Web, AI
- Stack
- RAG
- Gemini
- GrapesJS
- Python
- Vertex AI
- Flask
- JavaScript
- Prompt engineering
- Google ADK
Situation
Summer internship at Sofrecom Tunisia as an AI intern (July – August 2025), on an innovative project combining generative AI (RAG) and low-code interface generation.
The goal: go from a single sentence to an interface ready to edit, to speed up low-code application development.
What this internship gave me:
- designing and implementing a RAG system end to end;
- applying low-code technologies to speed up interface creation;
- contributing to an innovative project in a collaborative, professional environment.
Product
- The user describes the interface they want, in natural language.
- The RAG system retrieves the context relevant to the request, then generates the interface in GrapesJS, the low-code editor.
- The result stays visually editable: you start from a generated base instead of a blank page.
System
- RAG pipeline: retrieval of the relevant context, then generation by Gemini through Vertex AI.
- Prompt templates for structured, reliable outputs.
- Flask back end, GrapesJS / JavaScript front end, REST APIs.