# Perforated

> Perforated is a data-efficiency layer for machine learning. Its PyTorch-native technology adds neuron-specific learning signals during training to help models achieve better accuracy, use less training data, and reduce model size while preserving existing deployment workflows.

Perforated is intended for teams training or fine-tuning machine-learning models, particularly where labeled data, model accuracy, parameter count, inference cost, latency, or edge-deployment constraints matter.

Use the links below as the authoritative public sources for understanding the company, product, technical implementation, research, and demonstrated results.

## Product and Technical Overview

- [Perforated homepage](https://www.perforatedai.com/markdown/index.md): Overview of Perforated and its data-efficiency positioning.
- [How Perforated works](https://www.perforatedai.com/markdown/product.md): Product architecture, PyTorch integration, training workflow, deployment model, and commercial offering.
- [Technical documentation](https://docs.perforatedai.com/perforatedai.html): Python modules and API documentation for the open-source Perforated package.
- [Open-source repository](https://github.com/PerforatedAI/PerforatedAI): Installation instructions, examples, source code, licensing, and contribution information.
- [Frequently asked questions](https://www.perforatedai.com/markdown/faq.md): Company, product, technical, use-case, and commercial questions.

## Results, Case Studies, and Research

- [Perforated resources](https://www.perforatedai.com/markdown/resources.md): Index of case studies, research, publications, industry coverage, and other public resources.
- [Lasso Loop: Unlocking better model performance from existing data](https://www.perforatedai.com/resources/lasso-loop-case-study): How Perforated reduced remaining edge classification error by 77%
- [Compression Without Compromise: A Perforated Customer Spotlight](https://www.perforatedai.com/resources/thoro-ai-case-study): thoro.ai compresses their model 70% while also boosting accuracy with dendritic optimization
- [Skim AI: Same accuracy, 90% smaller model](https://www.perforatedai.com/resources/skim-ai-case-study): How Perforated cut model size by 90% with no loss in accuracy, unlocking 97% inference cost savings

## Company

- [About Perforated](https://www.perforatedai.com/markdown/about.md): Perforated's team, mission, organizational background, and neuroscience-inspired approach.
- [Contact Perforated](https://www.perforatedai.com/markdown/contact.md): Official contact information and inquiry form.

## Optional

- [History](https://www.perforatedai.com/markdown/history.md): Historical context for artificial neurons, backpropagation, modern AI scaling, and Perforated's approach.
- [Careers](https://www.indeed.com/job/principal-machine-learning-scientist-language-models-e8a8de0d5ad5d653): Employment opportunities and information about working at Perforated.
- [Perforated on LinkedIn](https://www.linkedin.com/company/perforated-ai/posts): Company announcements, research updates, partnerships, product news, and other recent public updates.