Every data source becomes a set of typed nodes in the knowledge graph. Edges between nodes encode relationships — co-authorship, topic overlap, proximity, endorsement. Scores flow through the graph to surface the most relevant connections.
Commit history, PR quality, repo topics, language distribution, contribution cadence.
Schema fields
Paper abstracts, citation graph, co-author relationships, topic embeddings.
Schema fields
Peer review quality, acceptance rates, venue prestige, reviewing consistency.
Schema fields
Technical discussion quality, topic focus, community engagement in ML/engineering subs.
Schema fields
Career trajectory, self-declared skills, education pedigree, tenure patterns.
Schema fields
Open slots for in-person meetups, preferred times, general location for proximity matching.
Schema fields
How nodes connect to form the knowledge graph.
Repo topics and languages map to skill nodes
Links paper nodes to their author profiles
Co-authorship creates high-weight collaboration edges
Commit diffs embedded to topic clusters
OpenReview peer assessment tied to paper nodes
Aggregated skill nodes anchor to person entity
Graph proximity score used for match ranking
Proximity edge for in-person meeting suggestions
Every skill score (0.00 – 1.00) is derived from graph traversal across all sources — frequency, recency, depth, and peer validation all feed into the final signal. No endorsement buttons. No keyword stuffing.
