FruitPunchSamuraiG/Llama3.2_Candidate_Evaluation_finetuned_3B

Llama3.2 Candidate Evaluation Finetuned (3B)

Repository: FruitPunchSamuraiG/Llama3.2_Candidate_Evaluation_finetuned_3B
Author: Hriday Ranka
Model Type: Finetuned LLaMA 3.2 (3B)
License: Apache 2.0

Model Summary

This model is a fine-tuned version of LLaMA 3.2 (3B), specifically adapted for contextual candidate-job matching. It takes in a resume and a job description and evaluates the semantic alignment between them, producing a relevance score along with detailed insights on match quality. Trained using Alpaca-style instruction tuning on a synthetically generated dataset of resume–JD pairs, the model captures deeper relationships such as skill transferability, domain overlap, and role suitability. This fine-tuned LLaMA model powers the core matching engine behind EmbedMatch, enabling scalable, explainable, and human-aligned candidate ranking decisions.

Training Details

• Base Model: LLaMA 3.2 (3B)
• Dataset: Custom dataset for job description generation
• Fine-tuning Method: Instruction tuning (Alpaca format)
• Quantization: Available in Q8_0, F16, Q4_K_M

Usage

Using llama.cpp

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
make

./main -m /path/to/unsloth.Q8_0.gguf -p "Generate a job description for a data scientist."