A multi-agent job intelligence system that combines semantic vector search, ML-powered ATS scoring, and a LangChain ReAct agent to find and rank the best-fit roles for a specific resume. Built for the LA Silicon Beach market. Personalizes every search to your actual skills, salary targets, and commute tolerance — not just keywords.
What if job search worked like a personal recruiter instead of a keyword filter? This is that.
Tech Stack
Features
Semantic Vector Search
SBERT sentence-transformer embeddings + ChromaDB. Finds roles by meaning, not keywords. Your resume becomes the query.
ATS Classifier
scikit-learn ML model trained on real posting data. Predicts pass rate with 96%+ accuracy and shows which keywords move the needle.
ReAct Agent Orchestration
LangChain multi-agent with autonomous reasoning. Uses 14 parallel MCP tools to filter, rank, and compare jobs — no human loop.
Resume-Aware Personalization
Auto-loads resume.json. Every search pre-filled with your skills, target roles, and salary preference. One-click matching.
Commute-Aware Scoring
LA Silicon Beach geospatial focus. Commute distance is part of the ranking model — not filtered out after the fact.
70× Cheaper Than GPT-4
DeepSeek as the reasoning backbone: $0.14/$0.28 per 1M tokens. Full agent intelligence at 1% of the typical LLM cost.