Hybrid AI Talent Intelligence Platform
Designed and shipped a hybrid talent-matching engine for Sedha Consulting that ranks candidates against open roles using a Three-Tier Ranking Algorithm — cosine similarity over spaCy en_core_web_lg 300-dimensional word vectors, keyword skill bonuses (+4%), and industry boosts (+5%) to rank 2,400+ resumes.
Built an ETL pipeline to normalize unstructured resumes (PDF, DOCX) into structured profile vectors, with Llama 3.1 + Groq used for skill extraction and JD summarization. SVM classifiers refine candidate-role fit before final ranking.
Built an interactive Streamlit dashboard with skill-density charts, PDF/TXT job-description parsing, and side-by-side candidate comparison. Caching reduced resource-heavy NLP model loading from 30s to under 1 second.
Semantic matching. A human decision at the end.
AI-EDITED EDITORIAL · ILLUSTRATIVE SCENE, NOT PROJECT DOCUMENTATION