README.md

---
title: Job Description Parser
emoji: 🧠
colorFrom: blue
colorTo: purple
sdk: streamlit
sdk_version: 1.48.0
app_file: app.py
pinned: false
---

# Job Parser Model (Qwen Fine-Tuned)

This repository contains a fine-tuned version of the [Qwen](https://huggingface.co/Qwen) model, specifically adapted to parse job descriptions into structured JSON format.

---

## 💼 Use Case

The model takes raw job descriptions (JD) as input and outputs structured JSON data containing:

- Job Titles
- Company Name & Website
- Skills
- Compensation
- Location
- Work Mode
- Experience
- Qualification
- Industry
- Posted Date
- Notice Period
- Job Type

Perfect for building:
- Resume & JD analyzers
- Job boards with smart filtering
- HR automation tools
- Job matching engines

---

## 🧠 Model Details

- **Base Model**: `Qwen` (`qwen/Qwen3-0.6B`)
- **Fine-tuned on**: 80+ custom-labeled job descriptions
- **Trained using**: Hugging Face Transformers & TRL's SFTTrainer
- **Dataset Format**: Few-shot prompting with Qwen’s `<|im_start|>` / `<|im_end|>` chat template
- **Output**: Structured JSON response

---

## 🚀 How to Use

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "Rithankoushik/job-parser-model-qwen"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16)

prompt = """<|im_start|>system
You are a helpful assistant that extracts structured information from job descriptions.
<|im_end|>
<|im_start|>user
[Paste job description here]
<|im_end|>
<|im_start|>assistant
"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=1024)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))