llm-profile-matcher

AI Profile Matching Agent 🧠✨

This project implements an intelligent AI agent that matches user-defined queries with the most relevant professionals based on structured profile data and unstructured LinkedIn-style resumes.

It was developed as part of the AI Agent Expert Matching Competition 2025.

🔍 What it Does

The agent analyzes a user's query (e.g. "Looking for a senior AI engineer with edge computing experience in Chicago") and scores the relevance of 40 mock professional profiles using a large language model (Gemini by Google).

It returns the top-matching candidates, including:

• Name and job title
• Location
• Contact info
• Availability
• Relevance score (0–10)
• AI-generated explanation of the match

---

🧭 Two Approaches Implemented

1️⃣ Dynamic Prompting (LLM-only mode)

• Sends the full query, Excel metadata, and HTML resume to Gemini
• Gemini handles all reasoning: skills, availability, location
• Best for open-ended, flexible matching

2️⃣ Filtered + Prompting (Hybrid mode)

• Applies dropdown filters (City, Job Title) to limit the candidate pool
• Sends each candidate’s structured + unstructured data to Gemini for scoring
• Balances precision and performance (quota-safe)

---

🚀 Run in Google Colab

To try it yourself, open the notebook below in Google Colab:

---

📁 Project Structure

• mock_profiles_aicompetition.xlsx: Excel sheet with name, job title, city, contact info, and availability
• /profiles/*.html: Mock LinkedIn pages for each candidate
• *.ipynb: Notebooks for both matching modes

---

🧑‍💻 Developed By

Sahar Zargarzadeh
PhD Student | Machine Learning Researcher
Built for the AI Matching Agent Challenge 2025

---