Reading the profile, not the keywords — a hybrid, CPU-only ranker that surfaces the top-100 candidates for a Senior AI Engineer JD, beating keyword-stuffers and honeypots. Structured JD-fit + contrastive static embeddings + behavioral/honeypot gating. No LLM at inference; reproducible in ~3 min.
python nlp machine-learning information-retrieval scikit-learn recsys semantic-search explainable-ai contrastive-learning cpu-only model2vec candidate-ranking static-embedding learning-ro-rank
-
Updated
Jun 28, 2026 - Python