bilalchawdhary/clinic_flow

🏥 ClinicFlow AI

A medical triage and patient intake assistant fine-tuned on Llama 3.2 3B using LoRA adapters.

---

📺 Demo Video

---

🚀 Live Demo

👉 Try ClinicFlow AI on HuggingFace Spaces

---

📌 What is ClinicFlow AI

ClinicFlow AI is a responsible clinic triage and patient intake assistant. It conducts multi-turn conversations to collect patient symptoms, asks focused follow-up questions, and generates a structured clinical summary at the end.

---

🔧 Model Details

Detail	Info
Base Model	unsloth/Llama-3.2-3B-Instruct
Fine-tuning Method	LoRA (PEFT)
Training Loss	0.107
Task	Medical Triage & Patient Intake
Language	English
License	MIT

---

🛠️ Tech Stack

• Python + PyTorch
• HuggingFace Transformers
• TRL 1.3.0 + PEFT + LoRA
• Trained on Google Colab A100 GPU
• Gradio UI

---

💬 How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

base_model_name = "unsloth/Llama-3.2-3B-Instruct"
adapter_name = "bilalchawdhary/clinic_flow"

tokenizer = AutoTokenizer.from_pretrained(base_model_name)

base_model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

model = PeftModel.from_pretrained(base_model, adapter_name)

prompt = """<|begin_of_text|><|start_header_id|>system<|end_header_id|>

You are ClinicFlow AI, a responsible clinic triage
and patient intake assistant.<|eot_id|>
<|start_header_id|>user<|end_header_id|>

I have a headache and fever since 2 days<|eot_id|>
<|start_header_id|>assistant<|end_header_id|>

"""

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
input_length = inputs["input_ids"].shape[1]

outputs = model.generate(
    **inputs,
    max_new_tokens=300,
    temperature=0.7,
    top_p=0.9,
    repetition_penalty=1.2,
    do_sample=True,
    pad_token_id=tokenizer.eos_token_id,
)

new_tokens = outputs[0][input_length:]
response = tokenizer.decode(new_tokens, skip_special_tokens=True)
print(response)

---

✅ Capabilities

• Multi-turn patient intake conversations
• Focused symptom follow-up questions
• Never diagnoses or prescribes
• Structured clinical summary at end
• Recommends consulting a real doctor

---

⚠️ Disclaimer

ClinicFlow AI is for demonstration purposes only. It is NOT a replacement for a real doctor. Always consult a qualified medical professional for diagnosis and treatment.

---

👨‍💻 Built By

Bilal Chawdhary — Software Engineering Graduate building AI products and sharing the journey.

• 🎥 Demo Video
• 🤗 Live Demo