sfisohlatshwayo/pathfinder-ai

PathFinder AI is an AI-powered career mentor designed for African students and professionals. It creates personalized career roadmaps from education to employment using Gemini, multi-agent systems, and real-time data. It recommends learning paths, bursaries, projects, and job opportunities tailored to local African contexts and opportunities. # Pathfinder AI A multi-agent **career guidance** system built with Google's Agent Development Kit (ADK). Give it your background and a target role, and a team of specialized agents will assess your skills, build a learning roadmap, find funding, plan a portfolio, and prepare you for interviews — then synthesize everything into one plan. It uses ADK `Agent` definitions, `SequentialAgent`/`ParallelAgent` orchestration, and a save-output-to-state callback so agents hand work to each other through shared session state. ## Core Features - **End-to-End Orchestration**: Seamlessly transitions from profile discovery to interview prep. - **Evidence-Based Assessment**: Uses optional GitHub integration to validate technical skills. - **Distributed Tools**: Leverages the Model Context Protocol (MCP) for out-of-process tool execution. - **Graceful Degradation**: External API dependencies (YouTube, Search) are optional and fail silently with best-effort advice. - **Privacy First**: Integrated PII filtering and explicit consent management for sensitive data. ## Architecture ``` pathfinder-ai/ ├── agents/ │ ├── orchestrator.py # SequentialAgent + ParallelAgent pipeline + synthesizer │ ├── discovery_agent.py # → user_profile │ ├── skills_agent.py # → skill_assessment (career_skill, github_skill) │ ├── roadmap_agent.py # → roadmap │ ├── scholarship_agent.py # → scholarships │ ├── portfolio_agent.py # → portfolio_plan (github_skill) │ └── interview_agent.py # → interview_prep (interview_skill) ├── skills/ # Plain-function tools, passed to agents as ADK FunctionTools │ ├── career_skill.py # role requirements, skill-gap comparison │ ├── github_skill.py # GitHub profile analysis, portfolio ideas │ └── interview_skill.py # question generation + answer rubric ├── mcp/ # Optional MCP servers (distributed tool sources) │ ├── github_server.py │ ├── youtube_server.py │ └─ …