Designed a matching workflow that treats profile and role data as semantic objects, then surfaces stronger matches through a product-friendly onboarding flow.
Case Study / Production client work
SteelCareer
Job discovery usually relies on brittle keyword matching, forcing candidates and recruiters through noisy search flows.
Semantic search and matching layer built into a Next.js product with Supabase-backed data and TypeScript application logic.
Created a clearer path from candidate signals to relevant opportunities without claiming unsupported performance metrics.
Technical highlights
What to inspect.
01
Modeled candidate and role attributes for retrieval-oriented matching.
02
Connected ML-facing ranking logic to a usable product flow.
03
Kept the system explainable enough for recruiter and candidate workflows.