Secure AI systems for long-term reliability.

Research and engineering focused on:

• AI security
• Prompt injection defense
• Reasoning systems
• Developer infrastructure

We build production-grade tools designed for safety, clarity, and maintainability.

OpenClay

Secure First → Execute Second

Universal zero-trust execution framework for LLM agents.

Repository: https://github.com/openclay-ai/openclay

prompt-injection-deberta

Security-focused DeBERTa model for prompt injection detection.

Model: https://huggingface.co/neuralchemy/prompt-injection-deberta

Prompt-injection-dataset

• 6k curated samples
• 15k augmented samples

Dataset: https://huggingface.co/datasets/neuralchemy/Prompt-injection-dataset

Website: https://www.neuralchemy.in
Hugging Face: https://huggingface.co/neuralchemy
GitHub Org: https://github.com/openclay-ai

Build systems that remain secure under real conditions.