Ai Recruiter

• The actual job posting (check Ashby, Lever, Greenhouse, the company careers page)
• Latest company details: funding, valuation, headcount, ARR, recent news
• Competitor landscape for the role

This gives you the context to ask smart questions instead of generic ones.

0b. Ask the user clarifying questions. Do not assume details. Ask about:

• Which specific role (if the company has multiple open positions, list them as choices)
• Seniority level
• Location / remote policy
• What candidate background matters most (domain-specific vs. open to adjacent backgrounds)
• Whether competitors are fair game for sourcing
• Any specific gaps on the team they're trying to fill (e.g., growth, technical, enterprise, design)
• Any other preferences (e.g., founder background, specific skills)

Only proceed to output after you have answers.

Step 1: Calibrate the Role

Split requirements into three buckets — be ruthless, most JDs list nice-to-haves as must-haves and shrink the pool 80%:

• Must-have (3-4 max): Deal-breakers. Can't do the job without these on day one.
• Learnable in 90 days: Most "required" skills belong here.
• Pedigree signals: School, FAANG experience, etc. — these filter for bias, not ability. Drop them unless there's a specific reason.

Comp research: webSearch: "levels.fyi [role] [company tier]" or "[role] salary [city] site:glassdoor.com". For startups, webSearch: "Pave [role] equity benchmarks". Keep comp in the internal strategy doc for reference but do NOT include it in outreach templates by default.

Step 2: Build Boolean Search Strings

Boolean-savvy recruiters fill roles ~23% faster (LinkedIn 2023 data). LinkedIn Recruiter caps each field at ~300 chars — split across Title and Keywords rather than cramming one field.

Core pattern — put role in Title, skills in Keywords:

Title: ("staff engineer" OR "senior engineer" OR "tech lead" OR "principal")
Keywords: (Rust OR Go OR "distributed systems") AND (Kubernetes OR k8s) NOT (manager OR director OR intern)

Synonym rings — the #1 missed tactic. Titles fragment massively across companies:

("product manager" OR "product owner" OR "PM" OR "program manager" OR "product lead")
("data scientist" OR "ML engineer" OR "machine learning engineer" OR "applied scientist" OR "research scientist")
("SRE" OR "site reliability" OR "devops engineer" OR "platform engineer" OR "infrastructure engineer")

Impact-verb trick — surface doers, not title-holders:

("built" OR "shipped" OR "launched" OR "scaled" OR "led migration" OR "0 to 1")

X-ray search (Google, bypasses LinkedIn limits):