Update README.md

README.md

@@ -1,3 +1,74 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+language:
+- en
+base_model:
+- Qwen/Qwen3-0.6B
+tags:
+- job-parsing
+- qwen3
+- lora
+- json-extraction
+---
+
+# 📦 Qwen3-0.6B — Job Description Struct-Extractor
+
+A fine-tuned version of **Qwen3-0.6B** designed for accurate extraction of structured job attributes from raw job descriptions.  
+This model outputs strict, schema-aligned JSON, making it perfect for downstream applications like search, analytics, and recommendation systems.
+
+---
+
+## 🚀 Model Highlights
+
+- **Base Model:** Qwen/Qwen3-0.6B  
+- **Architecture:** Decoder-only Transformer (Causal LM)  
+- **Tokenizer:** QwenTokenizer (same as base)  
+
+**Fine-Tuned For:**
+- Zero-hallucination extraction  
+- Schema-conformant JSON outputs  
+
+---
+
+## 🎯 Task Overview
+
+- **Task:** Extract structured fields from job descriptions  
+- **Output:** JSON strictly following a predefined schema  
+
+**Use Cases:**
+- Automated JD parsing into structured fields  
+- Talent platform search & recommendation engines  
+- HR data cleaning & analytics pipelines  
+- Resume ↔ Job matching systems  
+
+---
+
+## 🖥️ Inference Example (Python)
+
+```python
+import torch
+import re
+import time
+import json
+import json5
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from peft import PeftModel
+
+# Model paths
+base_model_id = "Qwen/Qwen3-0.6B"
+lora_model_id = "Rithankoushik/Qwen-0.6-Job-parser-Model"
+
+# Load tokenizer
+tokenizer = AutoTokenizer.from_pretrained(base_model_id, trust_remote_code=True)
+tokenizer.pad_token = tokenizer.eos_token
+
+# Load model + LoRA
+base_model = AutoModelForCausalLM.from_pretrained(
+    base_model_id,
+    trust_remote_code=True,
+    torch_dtype=torch.float16,
+    device_map="auto"
+)
+model = PeftModel.from_pretrained(base_model, lora_model_id, device_map="auto")
+model = model.merge_and_unload()
+model.eval()