EA Connect 2025 Scry

About

A networking tool for EA Connect 2025. Uses vector embeddings and LLM reranking to help you find the right people among 5,224 attendees (1,786 with rich profiles).

• Semantic search - "funders interested in AI safety", "people hiring ML engineers"
• HyDE - Generates hypothetical matching profiles for better results
• LLM reranking - Scores candidates on actual relevance
• Filters - By funding, mentorship, country, expertise

Download

# Or run manually:
uv venv && source .venv/bin/activate
uv pip install sqlite-vec requests

python ea_connect.py search "funders interested in AI safety"
python ea_connect.py find --interest "biosecurity" --funding
python ea_connect.py stats

Schema

-- SQLite with sqlite-vec for vector search

CREATE TABLE attendees (
    id TEXT PRIMARY KEY,
    name TEXT, role TEXT, organization TEXT,
    about TEXT, event_goals TEXT, can_help_with TEXT,
    country TEXT, has_funding BOOLEAN,
    is_mentor BOOLEAN, is_speaker BOOLEAN,
    seeking_collaborators BOOLEAN, profile_url TEXT
);

CREATE TABLE interests (attendee_id TEXT, interest TEXT);
CREATE TABLE expertise (attendee_id TEXT, skill TEXT);

CREATE VIRTUAL TABLE vec_embeddings USING vec0(
    attendee_id TEXT PRIMARY KEY,
    embedding FLOAT[3072]
);