Impact
A major US travel healthcare staffing firm posts thousands of jobs every day against a pool of more than 100,000 traveling clinicians. Since 2020, candidates have been shopping across agencies, so getting the match right the first time decides whether an order gets filled at all. Recruiters were searching the database by hand, which caused delays, missed opportunities and poor matches. Blackstraw built an AI matching platform that scores candidates against open jobs and reaches out to the best fits automatically. It optimized the matching process 5x, cut hiring cycle time by up to 30%, and reduced manual search effort by about half.
Background
A good match depends on more than discipline and specialty. Availability, license requirements, shift preferences and willingness to float to other units all matter. Many key hospital requirements were written as free-text notes instead of structured fields, so the old matching process never saw which criteria were required, preferred or simply nice to have. The client needed a system that could read those requirements, weigh every constraint and get the right jobs in front of the right clinicians quickly.
Solution Highlights
- Structured Requirements from Free Text: A named entity recognition model pulls experience, discipline, specialty, subspecialty and license requirements out of free-text notes and feeds them to the matching models.
- Hybrid Matching Engine: NLP for semantic matching, machine learning for classification and recommendation, and graph neural networks to improve match quality, combined through ensemble methods.
- Preference-Driven Outreach: A preference hub captures clinician and recruiter inputs. Matched clinicians receive push notifications in three rounds (immediately, then 24 and 48 hours later), and their interest goes straight to recruiters.
- Human-in-the-Loop Learning: Recruiter feedback flows back into the models, and a recruiting console and recommended-jobs view serve both sides.
Key Benefits
- 5x Optimized Matching: The job-to-candidate matching process was optimized 5x.
- Up to 30% Shorter Hiring Cycles: Hiring cycle time dropped by up to 30%.
- Hidden Requirements Captured: Critical free-text criteria reached the matching models as structured data for the first time.
- 50% Less Manual Search Effort: Recruiters spent about half as much time searching the database and more time with qualified candidates.
