GETFINALOFFER

Beyond the keyword.Closer to the right match.

Recruiting intelligence

Full-stack engineering

Live · Open source

Art-directed interface study · sample content

GetFinalOffer connects candidate profiles and verified interview progress to a recruiter’s search. A skill graph helps bridge the gap between the language of a role and the experience on a resume.

The problem

A recruiter may ask for a machine-learning engineer while a candidate lists Python, PyTorch, Docker, and MLOps. Exact keywords miss the relationship. Meanwhile, candidates need control over which organizations can discover their profiles.

The approach

I built a two-sided platform with asynchronous resume ingestion and tiered hybrid search. A taxonomy graph expands role and skill intent, reranking refines the results, and relevance notes give recruiters a reason to inspect a match.

Describe who you need.

Recruiter search accepts conversational queries and filters. Graph expansion connects role, skill, and alias relationships before the candidate list is ranked.

A match needs a reason.

Candidate cards surface skills, experience, verification, and relevance context. The profile preview brings the supporting details into the same review flow.

Keep the next step in view.

The recruiter dashboard brings invitations and their responses together. Candidate-side controls and private verification records support a more deliberate recruiting workflow.

How it’s built.

Make expensive ingestion asynchronous.

BullMQ separates upload handling from skill extraction, chunking, embeddings, and index publication. Content hashes avoid repeating embedding work for unchanged text.

Search should survive a dependency failure.

Exact and semantic caches, the live index, and the vector path form successive retrieval options. Graph timeouts and circuit breakers preserve baseline search.

Visibility applies to every access path.

Organization blocking is enforced in retrieval and direct profile access. Verification files stay private, with short-lived review links and an append-only decision record.

Ingest

Resume → extraction queue

Expand

Role, skill and alias graph

Rank

Hybrid retrieval + reranking

Review

Profile, evidence and invitation

Built with
Next.jsTypeScriptPostgreSQLpgvectorNeo4jRedisBullMQBetter Auth

Interface walkthrough · sample data · no audio

Find experience a keyword would miss.

A working recruiting platform connecting resume ingestion, skill-aware search, profile review, and invitations.

Graph ranking is feature-flagged for measured rollout. Local benchmark results are environment-specific, so this case study does not present them as production-wide performance.

Next project ↗

CERTUS

Workspace intelligence