You can deploy this model and start earning money today!

Yuchiwang02/
DelaySentinel

Name: DelaySentinel
Availability: InStock

safetensors

llama

text-classification

meta-llama/Llama-3.2-1B

finetune

apache-2.0

Views

No views yet

Model Card for DelaySentinel: AI-Powered Logistics Delay Prediction

Model Details

Model Description

Developed by: Yuchi Wang
Model type: Instruction fine-tuned large language model (LLM) for binary classification
Language(s): English (structured prompts with business/logistics features)
License: Apache-2.0
Finetuned from model: meta-llama/Llama-3.2-1B-Instruct

This model, DelaySentinel, was fine-tuned to predict whether a logistics order will be delayed (1) or not (0) before shipment, using structured order-level features. The project demonstrates how instruction fine-tuning of LLMs can be applied to supply chain risk management.

Model Sources

Repository: [Hugging Face Model Repo]
Demo: Hugging Face Gradio Space (interactive CSV/Excel upload)

Uses

Direct Use

Pre-shipment logistics delay prediction
Business analytics demos for supply chain risk management
Educational showcase of LLM instruction fine-tuning on structured business data

Downstream Use

Adaptation to other supply chain KPIs (e.g., demand forecasting, lead-time prediction)
Further fine-tuning on proprietary logistics datasets

Out-of-Scope Use

Not intended for sensitive decision-making in live operations without validation
Not suitable for medical, legal, or financial advisory

Bias, Risks, and Limitations

Dataset comes from Kaggle (synthetic/aggregated logistics data), so real-world generalization may be limited.
Model outputs are strictly 0/1 and do not provide uncertainty estimates unless re-trained for probabilities.
Risk of data drift if deployed in real supply chains with different carriers/regions.

Recommendations:
Users should validate predictions against recent operational data before deployment in practice.

How to Get Started with the Model

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("yuchi/DelaySentinel")
tokenizer = AutoTokenizer.from_pretrained("yuchi/DelaySentinel")

system = "You are a supply-chain analyst. Output only 0 or 1: 1=Delay, 0=Not delay."
user = "order_id: 123\norigin_region: OH\ndest_region: CA\ncarrier: A1\nservice_level: ground\nweight_kg: 10.5\ndistance_km: 3500\nholiday_flag: 0"
prompt = f"<|system|>{system}\n<|user|>{user}\n<|assistant|>"

out = model.generate(**tokenizer(prompt, return_tensors="pt"), max_new_tokens=2)
print(tokenizer.decode(out[0], skip_special_tokens=True).split("<|assistant|>")[-1].strip())

Yuchiwang02/DelaySentinel