Context
Devridge was inspired by a simple problem: developers working alone often wish they could ask people in different roles for feedback while building a project. Instead of leaving that support to chance, the idea was to create an LLM-based service that could act as a bridge between solo developers and role-specific professional perspectives.
Problem
Cross-functional advice is valuable, but it is not always easy to access when someone is building independently. A developer may want UI or UX feedback, performance suggestions, or code quality input, yet generic chat responses are often too vague to be useful. The challenge was to make feedback contextual, role-aware, and constrained enough to stay within the intended perspective.
Implementation
The project centered on prompt engineering. Developers submitted a project introduction, their current situation, and a preferred language for feedback, and the system used that context to generate more tailored responses. I designed role-specific constraints so each perspective, such as frontend-oriented feedback, would remain aligned with its intended scope rather than drifting into generic or irrelevant advice. This turned the product into a structured interface for contextualized LLM feedback rather than a general-purpose chat tool.
Outcome
Devridge became a lightweight but concrete prototype showing that prompt design and constraint-based role guidance can improve the usefulness of LLM-generated feedback. It also gave me practical experience in turning an abstract collaboration problem into a user-facing AI product concept with clear interaction boundaries.