aronjose2005/genz-resume-strength-classifier

GENZ Resume Strength Classifier

A fine-tuned DistilBERT model for classifying resume bullet points as weak or strong.

Labels

• 0 = weak
• 1 = strong

Dataset

The workshop dataset contains 400 labelled resume bullets:

• 200 weak
• 200 strong
• Multiple technical and professional domains
• Difficulty levels: easy, medium, and hard

The dataset was created specifically for this educational workshop.

Training

• Base model: distilbert/distilbert-base-uncased
• Training examples: 320
• Held-out test examples: 80
• Epochs: 3
• Learning rate: 2e-5
• Batch size: 8
• Random seed: 42

Evaluation

Held-out test results:

• Accuracy: 97.5%
• Precision: 100%
• Recall: 95%
• F1: 97.44%

Intended Use

This model is intended for educational demonstration and experimentation with domain-specific fine-tuning of Transformer models.

It can be used to explore whether a resume bullet communicates measurable impact, action, and specificity.

Limitations

This is a small educational dataset and should not be treated as professional hiring advice or as an objective measure of candidate quality.

Model confidence is not a guarantee of correctness.

Workshop

GENZ Resume AI · Livewire Trichy · September 2026

Built as a hands-on demonstration of:

1. Pretrained model inference
2. Domain mismatch
3. Tokenization and model outputs
4. Dataset preparation
5. Fine-tuning
6. Evaluation
7. Hugging Face model publishing
8. Gradio deployment