Inspiration

BU has thousands of researchers and alumni, but there's no easy way to see who's connected to whom. Students looking for a mentor, a lab, or an industry contact usually rely on luck. We wanted to make those connections visible, with every one backed by a real public record.

What it does

6degrees maps how BU researchers and alumni are connected through the papers they write together and the grants they share.

Search a topic (like "dementia" or "medical imaging") to see the BU people working on it, with verified alumni closest to the center.
Search a person to see who they work with, and click any connection to see the paper or grant behind it.
"How are we connected?" finds the shortest chain of real collaborations between two people.
Profile cards show someone's BU degree, current position, citation stats, and links to their ORCID, OpenAlex and their own websites.

## How we built it

Data pipelines (Python): papers, co-authors and topics from OpenAlex and Semantic Scholar; grants from NIH RePORTER; BU degrees from public ORCID records and OpenBU theses; profile details from ORCID and OpenAlex. Each module writes to one shared JSON schema, and a merge step combines them into a single validated graph.
Spec-driven with Kiro: specs, steering rules and hooks kept four people building against the same data contract.
Frontend (React + Cytoscape): a people-first network colored by BU status, built on our Lovable-designed UI. The path finder is a breadth-first search over people, where two people are linked if they share a paper or grant:

$$ d(u, v) = \min {\, k : u = p_0 \sim p_1 \sim \dots \sim p_k = v \,}, \quad k \le 6 $$

## Challenges we ran into

A data source disappeared mid-build. The PatentsView patent API was retired in March 2026, so we built an importer for USPTO's bulk-file replacement and shipped NIH grants first.
Identity resolution. The same person shows up in different databases under different IDs. Name matching alone mixed people up, so we only merge records on a unique identifier like ORCID.
Keeping the graph readable. Our first view mixed papers, topics and people into one tangle. Redesigning it around people made it usable.
Respecting data rules. Some useful sites (BU Profiles, LinkedIn) don't allow automated access, so we link to them instead of scraping them.

## Accomplishments that we're proud of

About 6,300 people, papers, topics and grants in one graph that validates with zero errors, where every connection links to its source.
Verified BU alumni went from 4 to 94 by checking ORCID education records and 27,000+ OpenBU theses.
Property-based tests caught a real bug (IDs with a hidden trailing newline) that 81 example tests had missed.

## What we learned

"Missing is fine, made up is not." Writing this rule down first shaped every decision. For example, someone is only labeled a BU alumnus when a source states their degree, never just because they published from a BU address.
Unique identifiers beat clever matching. ORCID did more for accuracy than any name-matching logic.
Design for the question people actually ask. Students want "who should I talk to?", not a database of records.

## What's next for 6degrees

Patents: turn on the USPTO importer once we have an API key.
Beyond AI in healthcare: expand from our demo topic to all BU research areas.
Smarter identity matching: let people who are already merged also link to their NIH grant records.
Let people add themselves: user-submitted profiles, so alumni can claim and enrich their own entry.