💼 Salary Predictor

A full-stack salary prediction pipeline that integrates Python, SQLite, Julia, Go, SQL, and Bash. Built to demonstrate working knowledge of ML pipelines, API development, and multi-language system design — and to learn Julia and Go, two languages I picked up the same day I built this. The model uses a fixed random seed for more consistent predictions, but variations may still occur.

---

🚀 Features

• 🐍 Python – Ingests and populates salary data into an SQLite database.
• 🗃️ SQLite – Lightweight relational DB used for structured storage.
• 🧠 Julia (MLJ) – Trains a regression model to predict salary using job_title, experience_level, and location.
• 💻 Go – Serves a lightweight HTTP API with HTML form and JSON endpoint.
• 🌐 HTML – Clean UI served from a static folder via Go.
• 🖥️ Bash – End-to-end pipeline runner (run_pipeline.sh).
• 🎯 RMSE on test set: $43,664

---

🧠 Model Details

The model is a DecisionTreeRegressor from the MLJ.jl ecosystem. It was chosen for its simplicity, interpretability, and ability to handle categorical features without manual one-hot encoding.

Preprocessing:

• Converted job_title, experience_level, and location columns to categorical types in Julia.
• No feature scaling was required due to the nature of the tree-based model.

Hyperparameters:

• Used default parameters (max_depth = -1, etc.) to keep the model configuration lightweight.
• Random seed set for reproducibility, though slight variation in predictions may still occur.

---

🧮 Baseline Comparison

To benchmark model performance, a simple DecisionTreeRegressor was trained in Python using scikit-learn on the same dataset and features (job_title, experience_level, company_location).

Python (scikit-learn) baseline results:

• RMSE: $49,820
• MAE: $37,564
• R²: 0.37

The Julia model (MLJ.jl) achieved an RMSE of $43,664, indicating slightly better performance while maintaining cross-language consistency.

This baseline comparison validates the effectiveness of the Julia model and provides a familiar point of reference for Python developers.

---

📂 Project Structure

salary-predictor/
│
├── data/                  # Raw CSV from Kaggle
├── Julia/                 # Model training in Julia
├── go-api/                # Go web server and HTML UI
│   └── static/index.html
├── Screenshots/           # UI screenshots
├── init.sql               # DB schema
├── run_pipeline.sh        # End-to-end bash runner
├── salary.db              # SQLite database
└── utl.py                 # Python ETL script

---

🗃️ Database Schema

Table: predictions

Column	Type	Description
job_title	TEXT	Role title
experience_level	TEXT	Entry / Mid / Senior
location	TEXT	Country or region code
predicted_salary	INTEGER	Model-predicted salary

---

🧪 Try It Locally

1. Clone and run:

git clone https://github.com/daniel-mehta/Salary-Predictor.git
cd salary-predictor
chmod +x run_pipeline.sh
./run_pipeline.sh

2. Open the browser:

Go to http://localhost:8080 Fill the dropdowns and click Predict Salary.

---

📊 Dataset

Kaggle – Data Science Salaries

---

🛠️ Built With

Language	Purpose
Python	Data ingestion & DB fill
SQL	DB schema/querying
Julia	ML model training
Go	API + web frontend
Bash	Full pipeline automation
HTML/CSS	User interface (form)

---

📄 License

MIT — free to use, modify, or extend.