SANYAMWADHWA.TECH :: PROJECT :: VERIDEX --- SUMMARY ---TITLE............. Veridex — Trap-Resistant Candidate Ranking EngineDATE.............. June 2026CATEGORY.......... ML & ResearchAWARD............. Redrob Data & AI ChallengePROBLEM........... Embedding-based resume screening collapses on adversarial data: keyword stuffers and impossible profiles float to the top while plain-language experts sink. Veridex grades impossibility, penalizes unearned skill claims, and still ranks 100K candidates in 34 seconds.WHAT IT DOES...... Candidate ranking engine that scores 100,000 resumes against a job description in 34 seconds on CPU. Hybrid BM25 + FAISS retrieval, 56 deterministic signals, and honeypot detection built to survive adversarially constructed datasets. --- RESULT ---METRIC 1.......... 100K candidates in 34s (CPU)METRIC 2.......... 0 honeypot false positivesMETRIC 3.......... 8/8 evaluation checks passing --- WHAT I BUILT ---HIGHLIGHT 1....... Two-layer scoring: JD-agnostic universal signals (60%) plus role-specific intent-graph matching (40%), modulated by confidence, availability, and graded honeypot and disqualifier factors.HIGHLIGHT 2....... Zero honeypot false positives in the top-100 while catching 3,492 traps pool-wide, with 100% ranking stability across 50 trials of ±10% weight perturbation. --- HOW IT WORKS ---STEP 1............ React + Vite DashboardSTEP 2............ FastAPISTEP 3............ BM25 + FAISS Hybrid Shortlist (k=3,000)STEP 4............ 56-Signal Feature ExtractionSTEP 5............ Honeypot Audit (L0–L3)STEP 6............ Two-Layer Scoring (60% universal + 40% JD)STEP 7............ Elite Re-scoreSTEP 8............ MMR Diversity RerankSTEP 9............ Cited Reasoning + CSV --- FEATURES ---FEATURE 1......... Detects four trap classes: keyword stuffers, behavioral twins, plain-language experts, and impossible profilesFEATURE 2......... Soft multiplicative penalties; only provably impossible profiles hard-zero, everything else degrades gracefullyFEATURE 3......... Top-50 elite pool re-scored with NDCG@10-tuned head weights, then MMR reranking (λ=0.7) for diversityFEATURE 4......... Evidence-grounded scoring: every number traces to real record fields with Role#N citations, so hallucinated reasoning is structurally impossibleFEATURE 5......... Fully config-driven; swapping one rubric.yaml adapts the entire engine to a new role with no code changesFEATURE 6......... Dashboard with live ranking, weight sliders, stability badges, fairness audit, and candidate radar comparisons --- STACK ---STACK............. Python, FastAPI, FAISS, BM25, sentence-transformers, React --- LINKS ---CODE.............. https://github.com/SanyamWadhwa07/Redrob-RankerHUMAN PAGE........ https://sanyamwadhwa.tech/projects/veridex