🌍 Job Market Intelligence & Career Advisor

Ironhack Final Project — Data Science & Machine Learning

A dashboard that answers "How much should I earn, and where are the jobs?" It combines salary analysis (EDA), salary prediction (Machine Learning), an AI career coach (RAG chatbot), and a live job finder across 7 regions.

---

🚀 Quick start

1. Install everything

Open a terminal in this project folder and run:

bash setup.sh

This creates a private environment (.venv) and installs all the libraries. It automatically installs the CPU-only version of torch, so it works on any computer — with or without an NVIDIA GPU (most laptops have none).

Windows? Don't worry — run these three commands instead in PowerShell:

python -m venv .venv
.\.venv\Scripts\pip install torch --index-url https://download.pytorch.org/whl/cpu
.\.venv\Scripts\pip install -r requirements.txt

2. API keys & the AI Career Coach

Create a file named .env in this folder (same place as README.md).

The AI Career Coach runs on a local model by default — Qwen3-Coder-Next served by llama.cpp — so it needs no API key. If you prefer the cloud, you can switch it back to the DeepSeek API instead:

# AI Career Coach — pick ONE option:
#   A) local model (llama.cpp, no real key — "local" is just a placeholder):
DEEPSEEK_API_KEY=local
DEEPSEEK_BASE_URL=http://localhost:8080/v1
DEEPSEEK_MODEL=qwen3-coder-next
#   B) cloud DeepSeek API (uncomment to use instead):
# DEEPSEEK_API_KEY=sk-...                  # → https://platform.deepseek.com
# DEEPSEEK_BASE_URL=https://api.deepseek.com
# DEEPSEEK_MODEL=deepseek-chat

TAVILY_API_KEY=tvly-...     # for the Live Job Finder   → https://tavily.com

To use option A, run your local model with llama.cpp's llama-server on http://localhost:8080 and keep it running while you use Tab 3. ⚠️ Never share this file — it holds your private keys.

3. Start the app

.venv/bin/streamlit run app/streamlit_app.py

Your browser opens http://localhost:8501 with the dashboard. That's it! 🎉

---

📁 Project layout

Folder / file	What it is
app/	⭐ The dashboard — streamlit_app.py is the app you run
src/	The "engine room" — Python modules the app uses
notebooks/	Step-by-step Jupyter notebooks (explore → ML → RAG)
data/	The data: raw/ (downloaded) and processed/ (cleaned)
sql/	Database schemas (how jobs/applications are stored)
visuals/	Chart-export tool + Tableau guide (optional extras)
models/	Trained ML models (created when you run the notebooks)
setup.sh	One-command installer (Linux/macOS)
requirements.txt	List of libraries to install
.env	Your secret API keys + optional local-model settings (you create this)
.venv/	The private environment (created by setup.sh — leave it)

What each src/ module does

Module	Job
data_loader.py	Reads & cleans the salary data
preprocessing.py	Prepares data for machine learning
models.py	Trains & evaluates the ML models
rag_engine.py	Powers the AI Career Coach (retrieval + local Qwen3 or DeepSeek API)
tavily_job_finder.py	Live job search across 7 regions
applications.py	Saves jobs & sends applications (SQLite)
utils.py	Shared settings & API keys

---

🧠 The 4 tabs

Tab	What it does
📊 Market Explorer	Interactive charts — salary by experience, company size, field
💰 Salary Predictor	Pick role + experience + location → estimated salary
🤖 AI Career Coach	Chat with a RAG assistant backed by real salary data (local Qwen3 — no key needed; cloud DeepSeek optional)
🌍 Live Job Finder	Search live DS/ML/DL + Aerospace jobs in 7 regions (needs Tavily key)

---

📓 The notebooks

Run them if you want to re-generate the data, charts, and models from scratch:

#	Notebook	What it does
1	01_data_exploration.ipynb	Downloads data, cleans it, creates data/processed/*.csv + charts
2	02_ml_modeling.ipynb	Trains ML models, saves them to models/
3	03_rag_build.ipynb	Builds the ChromaDB vector store for the AI Coach

The app works out of the box because data/processed/*.csv is already included. The notebooks are there so you can see how it was built.

---

🛠 Tech stack

Layer	Technology
Data	Pandas, NumPy
Charts	Matplotlib, Seaborn, Plotly
ML	Scikit-learn, XGBoost, LightGBM
RAG / AI	Qwen3-Coder-Next (local via llama.cpp) or DeepSeek API, LangChain, ChromaDB, sentence-transformers
App	Streamlit
Live search	Tavily API

---

built by enthusiasts