YouTube Video
Project Description
Perfert CV is an AI-powered career navigator and gamified workplace simulator that moves far beyond traditional, static chat interfaces. By leveraging Generative UI, our agent doesn’t just reply with text—it renders dynamic, actionable React components directly in the conversation stream. For example, when analyzing a user’s career trajectory, the agent dynamically generates a real-time CareerDashboardUI featuring visual salary market estimates, interactive job matches, and clickable action items.
This specific implementation of generative UI solves the real-world problem of overwhelming career transitions by transforming dense, text-heavy advice into intuitive, bite-sized visual tools. Our working codebase demonstrates originality by fusing a gamified “workplace survival” simulation (testing soft skills like EQ and data analysis) with an actionable, GenUI-driven resume builder, offering a completely novel and highly interactive approach to professional development.
Technical Stack
Our application is built on a modern, robust frontend stack using React (via Vite ), Zustand for state management, and Tailwind CSS alongside Radix UI for accessible, responsive design.
To power the core interactive AI experience, we integrated CopilotKit ( @copilotkit/react-core , @copilotkit/react-ui ). Instead of standard text responses, CopilotKit enables our agent to seamlessly render custom Generative UI components (such as our CareerDashboardUI and ExperienceFormUI ) directly within the chat timeline.
On the backend, we utilize an Express.js server to host the CopilotKit runtime ( @copilotkit/runtime ). To achieve advanced reasoning capabilities, we integrated the DeepSeek API ( deepseek-chat ). Because we needed to route requests correctly to DeepSeek’s /v1/chat/completions endpoint, we implemented a custom OpenAIAdapter monkey-patch on the server side using the Vercel AI SDK ( @ai-sdk/openai ). This custom integration alongside CopilotKit ensures reliable generation of our dynamic UI components and complex simulation logic.
