I've always been curious: where do the world's most impactful open-source developers actually live? Which cities produce the most stars? Is the "Silicon Valley dominates everything" narrative still true?

So I built DevGlobe — an interactive 3D globe that maps 26,000+ top open-source developers to their real-world locations, scored and ranked by their actual contributions.

https://devglobe.vercel.app

How it works

Every developer is scored 0–100 using a composite algorithm across 6 dimensions:

All dimensions are log-normalized to prevent one mega-repo from dominating the scores.

The data pipeline

The data comes from three sources, stitched together by a Node.js pipeline:

1. GitHub GraphQL API — pulls top developers by follower count, then fetches their stars, commits, repos, and languages

2. StackOverflow API — enriches profiles with reputation, answer count, and badges

3. Geocoding — converts free-text locations ("SF", "Bangalore, India", "🇧🇷") into lat/lng coordinates

The final dataset is stored in Azure Cosmos DB with vector embeddings, enabling semantic search.

AI-powered search

This is the part I'm most proud of. You can type natural language queries like:

• "Python developer working on AI and deep learning"

• "Full stack JavaScript developer in London"

• "Linux kernel contributors"

It uses Azure OpenAI embeddings + Cosmos DB vector search to find semantically relevant developers — not just keyword matching.

The search supports three modes:

• Text — fast client-side filtering by name, username, location

• Vector — semantic similarity via embeddings

• Hybrid — combines both using Reciprocal Rank Fusion (RRF)

Interesting patterns I found

After staring at this globe for way too long, here's what stood out:

🌎 The US West Coast and Western Europe dominate, but clusters in India, China, and Brazil are growing rapidly

💻 C and JavaScript developers are the most geographically distributed

🏙️ The top 100 developers live in just ~12 cities — San Francisco, London, Beijing, Bangalore, Berlin, Tokyo, Shanghai, New York, Singapore, Sydney, Toronto, Portland

📊 StackOverflow reputation doesn't correlate strongly with GitHub stars — the top GitHub developers rarely answer SO questions, and top SO contributors often have modest repos

🇺🇦 Eastern Europe punches above its weight — Ukraine, Poland, and Czech Republic have disproportionately high scores per capita

Tech stack

Why Azure Cosmos DB for search?

I needed a database that could handle both traditional queries AND vector similarity search in a single query. Cosmos DB's native vector search meant I didn't need a separate vector database — the embeddings live alongside the developer data, and I can combine text filters with vector distance in one query.

The hybrid search uses Reciprocal Rank Fusion to merge results from both approaches:

// RRF fusion — combines vector relevance with text match ranking
const k = 60;
const rrf = new Map();
vectorResults.forEach((r, i) => {
  rrf.set(r.login, (rrf.get(r.login) || 0) + 1 / (k + i + 1));
});
textResults.forEach((r, i) => {
  rrf.set(r.login, (rrf.get(r.login) || 0) + 1 / (k + i + 1));
});

Try it yourself

🔗 Live demo: devglobe.vercel.app

🔗 GitHub: github.com/sajeetharan/devglobe

The project runs locally with zero configuration — no API keys needed:

git clone https://github.com/sajeetharan/devglobe.git
cd devglobe
npm install
npm run dev

It ships with sample data (20 developers) so you can explore the UI immediately. For the full database experience, there's a one-command script to seed the Cosmos DB Emulator locally:

npm run seed-emulator

What's next

• 🗺️ Country-specific landing pages (SEO for "top developers in X")

• 🏅 Embeddable profile badges for GitHub READMEs

• 📈 Time-series view showing how the developer landscape shifts over years

• 📦 More data sources (GitLab, npm downloads, conference talks)

Contributing

The project is fully open source and contributions are welcome. The CONTRIBUTING.md has all the details. Whether it's improving the scoring algorithm, adding data sources, or fixing UI bugs — PRs are appreciated.

If you find it useful, a ⭐ on the repo goes a long way!

Have questions or feedback? Drop a comment below or find me on GitHub and Twitter.