Pathwise

Pathwise: From career confusion to a concrete plan.

Comment

🌟 Pathwise: From Career Confusion to a Concrete Plan

💡 Inspiration

Career decisions are stressful. People spend hours researching and still feel uncertain. Pathwise turns that confusion into clarity—with data, evidence, and actionable steps.

🛠️ Run Instructions

✅ Prerequisites

  • Node.js 18+
  • MySQL or TiDB instance
  • API keys for Google Gemini and SerpAPI

📦 1. Install dependencies

npm install

⚙️ 2. Configure environment

Create a .env file with at least:

DATABASE_URL="mysql://user:pass@host:port/db"
AUTH_SECRET="dev_secret_change_me"
GOOGLE_API_KEY="your_gemini_key"
SERPAPI_KEY="your_serpapi_key"

# Optional: Google Calendar OAuth
GOOGLE_CLIENT_ID=""
GOOGLE_CLIENT_SECRET=""
GOOGLE_REDIRECT_URI=""
ENCRYPTION_SECRET=""

▶️ 3. Start the dev server

npm run dev

App runs at 👉 http://localhost:3000

🚀 What It Does

Pathwise helps you:

  • 🗺️ Map career paths from your current skills.
  • ⚖️ Compare options side-by-side (bootcamp vs. MBA, transfer vs. stay).
  • 🔍 Surface real evidence from “people like me.”
  • 📅 Generate a week-by-week plan you can sync to Google Calendar.
  • 💬 Chat with a context-aware career coach.

🛠️ How We Built It

  • Frontend: Next.js + React
  • LLMs: Google Gemini via LangChain
  • Database: TiDB Cloud + vector search
  • Evidence: SerpAPI + RAG for trustworthy citations
  • Integration: Google Calendar for actionable planning

🔄 Data Flow & Integrations

  • User Profile & Auth → JWT cookie; profile stored via Prisma.
  • Path Explorer → Uses skills/resume → predicts target roles → fetches requirements & resources → saves results.
  • Decision Duel → LLM compares strategies, highlights trade-offs, merges evidence.
  • People Like Me → Vector similarity surfaces anonymized professional journeys.
  • Week Plan → Creates a 12-week structured schedule; Gemini fills in details.
  • Chat → LangChain graph connects Gemini to tools (path, duel, people, plan).

🌟 Main Features

  • Path Explorer – Interactive skill-to-role graph with confidence bands.
  • Decision Duel – Compare strategies with compensation, timeline, and burnout risk.
  • Week-by-Week Plan – Adaptive 12-week plan with auto-adjustments.
  • “People Like Me” Receipts – Real anonymized trajectories for trust and relatability.

⚡ Challenges

  • Keeping LLM outputs structured and reliable.
  • Designing fallbacks to prevent stalls.
  • Smoothly integrating multiple external APIs.

🏆 Accomplishments

  • Built a full journey: resume → role paths → actionable plan.
  • Effective personalization with TiDB vector search.
  • Context-aware chat widget with live session memory.
  • Resilient system with fail-safes.

📚 Learnings

  • Vector search dramatically improves personalization.
  • Transparency and citations build user trust.
  • Calendar integration makes the product practical.
  • Strong onboarding boosts long-term success.

🔮 What’s Next

  • 🌍 Expand into more industries and roles.
  • 👥 Add community features: connect with peers.
  • 🎚️ Smarter customization for financial goals and risk tolerance.
  • 🎓 Partner with course providers for actionable follow-through.

✨ Pathwise helps you go from career confusion to a concrete plan—step by step.

Built With

Share this project:

Updates