SAFELINK AI

# Collaboration Opportunity: SafeLink AI
**Project:** Explainable Phishing and Scam Message Detector
**Duration:** 7–14 days
**Collaboration type:** Unpaid academic / portfolio collaboration
**Technology:** Python, Pandas, scikit-learn, Streamlit, Joblib, Git/GitHub

## 1. Project overview

I am developing SafeLink AI as a college group experiential learning project. The objective is to build a simple web application that analyses text messages, predicts whether a message belongs to the classes supported by the selected dataset, highlights observable warning indicators, and provides general safety guidance.

I already have a detailed project proposal explaining the objectives, scope, methodology, mathematical logic component, evaluation metrics, and expected deliverables.

I am looking for a developer who is interested in collaborating on the implementation and creating a working, well-documented prototype.

## 2. Required features

**A. Data preparation and machine learning**
- Use a documented, labelled message dataset appropriate to the classification task.
- Load and inspect the dataset using Pandas.
- Handle missing values, duplicate records, and labels appropriately.
- Split the data into training and test sets.
- Implement a scikit-learn pipeline using TF-IDF and Logistic Regression.
- Save and load the trained pipeline using Joblib.
- Record actual test-set accuracy, precision, recall, F1-score, and confusion matrix.

**B. Propositional logic module**
- Implement Boolean indicators for suspicious-looking URLs, sensitive-credential requests, urgent wording, and sender/request verification status.
- Demonstrate the decision rule: D = (p AND q) OR (r AND NOT s).
- Include a complete 16-row truth table.
- Display machine-learning predictions and rule-based indicators separately so users can understand the difference.

**C. Streamlit user interface**
- Project title and a brief introduction.
- Text area for entering a sample message.
- Analyse button and input validation.
- Predicted class and relevant warning indicators.
- General safety recommendations.
- Clear disclaimer that results are advisory and cannot guarantee a message is safe.

**D. Testing**
- Test empty input, ordinary messages, urgent wording, OTP-related wording, URLs, long text, and missing model files.
- Test with held-out examples that were not used during training.
- Record test outcomes and fix reproducible errors.
- Do not invent accuracy values or present untested features as complete.

## 3. Deliverables

1. Complete Python source code.
2. Streamlit application that runs locally.
3. Dataset source, attribution, and usage/licence information.
4. Training and inference scripts.
5. Saved model file or reproducible instructions for generating it.
6. `requirements.txt` and README with installation and execution instructions.
7. Evaluation metrics and confusion matrix generated from actual tests.
8. Truth table and explanation of the mathematical logic.
9. Basic test report and application screenshots.
10. Final code walkthrough and handover session.

## 4. Beginner-friendly requirements

- Use clear filenames, simple functions, meaningful variable names, and comments where useful.
- Avoid unnecessary frameworks, paid APIs, databases, Docker, or complicated deployment.
- Explain the purpose of each module and how the ML pipeline works.
- Keep the project easy to run on a standard Windows laptop.
- Do not store user-submitted messages by default or ask users to enter real passwords, OTPs, or banking information.

## 5. Proposed milestones

- Days 1–2: Scope confirmation, dataset selection, repository setup.
- Days 3–4: Data preparation, baseline model, initial scanner.
- Days 5–7: UI, Boolean rules, model integration.
- Days 8–10: Testing, evaluation, documentation.
- Days 11–14: Bug fixes, code walkthrough, final handover.

We can agree on smaller milestones depending on availability. Please communicate early if the timeline needs adjustment.

## 6. Collaboration expectations

This is an unpaid collaboration, not a paid freelance contract. I can offer appropriate contributor credit and a portfolio acknowledgement, subject to mutual agreement.

Before starting, we will agree on task ownership, the deadline, attribution, code licensing/reuse permissions, and what constitutes completion.

Please share a brief introduction, your relevant Python/Streamlit experience, and any similar project or GitHub repository you have worked on.

Thank you!