swaraj20/resume_gte_embedding

📂 Dataset

This model was fine-tuned on the Resume Dataset from Kaggle(https://www.kaggle.com/datasets/palaksood97/resume-dataset/discussion/174335), containing resumes categorized by job roles. The data was used to generate similarity pairs and synthetic job descriptions for training.

🧠 Resume GTE Embedding Model

A domain-specific sentence embedding model fine-tuned on resumes using GTE-base. Built for use in HR systems, resume search engines, talent matching, and AI-driven hiring.

---

📌 Model Card

• Base Model: thenlper/gte-base
• Fine-tuned on:
  • Self-supervised resume chunks
  • Resume pairs with similar job roles (category-based)
  • Resume ↔ synthetic job description pairs
• Dataset Size: 4,169+ semantic chunks
• Loss: CosineSimilarityLoss
• Vector Output Size: 768
• Chunking: Semantic section-based splitting (e.g., “Work Experience”, “Skills”, etc.)

---

🧪 Intended Use Cases

• ✅ Resume similarity search
• ✅ Resume ↔ Job description matching
• ✅ Candidate clustering by skills or experience
• ✅ Career path embeddings / recommendation

---

🚀 Example Usage (Python)

from sentence_transformers import SentenceTransformer, util

model = SentenceTransformer("swaraj20/resume_gte_embedding")

resume1 = "Experienced software engineer skilled in Python, ML, and cloud deployment."
resume2 = "Looking for a backend developer with Python and AWS experience."

embedding1 = model.encode(resume1, convert_to_tensor=True)
embedding2 = model.encode(resume2, convert_to_tensor=True)

similarity = util.cos_sim(embedding1, embedding2)
print(f"Similarity Score: {similarity.item():.4f}")