Pangea AI

Inspiration

During COVID-19, a virologist in Tokyo had the genomic data. A field epidemiologist in Senegal had the population access. A vaccinologist in Paris had the platform.

They never connected.

The cost of that silence: 7 million lives.

This wasn't a failure of science. It was a failure of infrastructure.

Today, 76% of vaccine clinical trial participants are White — while the communities bearing the highest disease burden in West Africa, South Asia, and Southeast Asia remain absent from the research designed to protect them.

A vaccine tested on part of humanity protects part of humanity.

The data that would make vaccines truly universal exists. It lives in the field — in Amazonian ecosystems, in Sahelian surveillance networks, in South Asian cohort registries.

It just never reaches the lab.

Louis Pasteur wrote in 1876: "Science knows no country, because knowledge belongs to humanity, and is the torch which illuminates the world."

In 2026, that torch still doesn't travel far enough.

Pangea AI was built to change that — inside Slack, where scientific collaboration already happens.

Not another platform to log into. Not another directory to browse.

An agent that works where scientists already are. That sees who needs whom. That makes the connection happen — before the next outbreak begins.

Because the next pandemic is already forming somewhere. And the researcher who can stop it may already know someone who knows someone who has exactly what they need.

They just don't know it yet.

What it does

Pangea AI is a Slack agent that connects vaccine researchers with complementary expertise and unique geographic field access.

• /pangea [topic] — detects your Slack identity and finds your ideal collaborator. Not a similarity score — a precise explanation of what YOU bring to THEM and what THEY bring to YOU, including geographic and ecosystem access that cannot be replicated elsewhere.

• Collaboration Network Graph — visualizes the existing network of past collaborations in real time. Detects shared connections between you and your suggested match, and surfaces warm introduction opportunities through mutual colleagues.

• AI-drafted introduction message — scientifically grounded, culturally adapted, referencing the recipient's recent work. Delivered directly to their Slack DM in one click.

• App Home dashboard — your global research network at a glance: researchers across 5 continents, their expertise, collaboration history, and live impact metrics.

• Proactive matching — Pangea AI detects research topics in channel conversations and surfaces collaboration opportunities automatically, without any command needed.

How I built it

• Slack Bolt for Python with Socket Mode
• Anthropic Claude Haiku for AI-generated scientific collaboration proposals
• NetworkX + Matplotlib for real-time collaboration network graph generation
• Slack Block Kit for rich interactive UI cards
• Slack users.info API for automatic researcher identity detection from Slack profiles
• In-memory impact counter tracking matches, countries, and disciplines in real time
• 5 researcher profiles across 5 continents, anchored on real Institut Pasteur network disciplines
• 12 past collaboration entries powering the network graph and shared connection detection

Challenges I ran into

The main technical challenge was building on Windows with Python virtual environments — the Slack CLI v4.4.0 has known compatibility issues with PowerShell 5.1 that required upgrading to PowerShell 7 and careful environment management.

Generating scientifically accurate collaboration proposals required careful prompt engineering. Claude Haiku with a scientific writing system prompt produced outputs grounded in real research profiles, culturally adapted to each researcher's geographic context.

Building the collaboration network graph with NetworkX and Matplotlib inside a dark-themed Slack workspace required significant styling work to achieve a professional result that matches Slack's aesthetic.

Accomplishments that I'm proud of

A full end-to-end scientific collaboration workflow — from topic detection to personalized DM delivery.

The collaboration network graph, combining NetworkX and Matplotlib with Slack's dark theme, provides a genuinely novel visualization of research relationships that no existing tool offers inside Slack.

The shared connection detection feature — identifying mutual colleagues between a user and their suggested match, and quantifying the network distance — transforms Pangea AI from a matching tool into a true network intelligence agent.

What I learned

Block Kit is more powerful than it looks for building professional research tools. The combination of Slack identity detection via user profiles and profile-based bidirectional matching creates a genuinely personalized experience that generic chatbots cannot replicate.

The most impactful features were the ones that combined Slack-native capabilities (identity, DMs, channels) with external intelligence (network graph, Claude-generated proposals) — neither alone would have been as compelling.

What's next for Pangea AI

• MCP server integration for live researcher profiles from ORCID and Semantic Scholar
• Real-Time Search API for proactive weekly match suggestions based on workspace activity
• Slack Marketplace submission for broader research community access
• Multi-workspace support for international research networks like the Institut Pasteur global network
• Integration with academic calendar APIs to suggest collaboration windows

Built With

• anthropic-claude
• block-kit
• matplotlib
• networkx
• python
• slack-bolt