Hassan El Mghari

El Mghari leads Together AI’s team for demonstrations, documentation, and content. He previously worked at Vercel and ran a game-marketing company. His examples connect model APIs with a complete interface, storage, and deployment code that other developers can inspect.[1]

A photo becomes a new room design

RoomGPT’s open-source version takes a room photo and sends it through ControlNet to produce variations. A Next.js API route connects the interface to the model, with Replicate hosting the model and Bytescale storing images. The repository identifies this as the first version, before the paid product’s authentication, payments, and additional features. It is a compact example of adapting an existing model to a specific user task.[4]

A design procedure for coding agents

Hallmark packages interface-design instructions, visual themes, and browser checks as a skill for coding agents. Its current README separates a basic quality floor, such as contrast and keyboard focus, from stylistic defaults and finish checks. It can build, audit, or redesign a page while checking that a redesign preserves its links, fields, and handlers. The repository is owned by El Mghari and explicitly credits Together AI, so it represents a team-associated project rather than a claim of sole authorship.[2][3]

In a first-person architecture post, El Mghari describes a recurring application pattern: accept text or an image, send it to a model with a task-specific prompt, store results when persistence is needed, and display them. He prefers a single model call when it can solve the problem, to shorten development and test an idea sooner. His broader portfolio applies this approach to photo restoration, room design, logo generation, and code generation.[1][5]