---
language:
- en
license: mit
tags:
- text-classification
- sequence-classification
- enterprise-operations
- logistics
- finops
- anomaly-detection
datasets:
- priteshloke/enterprise-operations-benchmark
pipeline_tag: text-classification
widget:
- text: "Carrier BlueDart billed 3.5 kg on a 0.5 kg t-shirt box due to dimensional laser scanner bulge."
  example_title: "Volumetric Freight Overcharge"
- text: "Client logged 68 hours on a contracted 40-hour monthly retainer with zero change orders."
  example_title: "Agency Retainer Scope Creep"
- text: "Departed marketing contractor retains paid Google Workspace and Slack licenses for 90 days."
  example_title: "SaaS Zombie Seat Waste"
---

# 🤖 Kepler Ops — Enterprise Anomaly & Exception Classifier

### Author: Kepler Operations Intelligence ([getkeplerops.com](https://www.getkeplerops.com))

This model classifies transaction narratives, invoice discrepancy notes, and operational logs into standard enterprise failure modes.

## 📊 Target Classes (6 Exception Categories)
1. `VOLUMETRIC_WEIGHT_OVERCHARGE_CRITICAL`
2. `COURIER_FAKE_NDR_ATTEMPT_CRITICAL`
3. `AGENCY_RETAINER_SCOPE_CREEP_CRITICAL`
4. `SAAS_DORMANT_SEAT_LICENSE_WASTE_HIGH`
5. `FBA_AGED_INVENTORY_SURCHARGE_CRITICAL`
6. `NONE_CLEAN_TRANSACTION`

## 💻 Quick Python Usage
```python
from transformers import pipeline

classifier = pipeline("text-classification", model="priteshloke/kepler-ops-anomaly-classifier")
result = classifier("Carrier billed 3.5 kg on a 0.5 kg shipment.")
print(result)
```

## 🚀 Live Production Engine & Benchmarks
- **Web Platform:** [getkeplerops.com](https://www.getkeplerops.com)
- **Interactive Simulator:** [Hugging Face Space](https://huggingface.co/spaces/priteshloke/kepler-ops-anomaly-playground)
- **Evaluation Dataset:** [priteshloke/enterprise-operations-benchmark](https://huggingface.co/datasets/priteshloke/enterprise-operations-benchmark)