Protopia provides Stained Glass technology to enable secure use of sensitive data within AI environments without compromising performance. Their Stained Glass Transform converts enterprise data into stochastic representations for privacy-preserving model training and deployment. This solution allows organizations to maintain data ownership while securely processing information across on-premise, hybrid, and cloud infrastructures.
Funding
$25M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
Organizations face challenges in securely accessing and sharing sensitive data for AI and machine learning initiatives due to the risk of data breaches and privacy violations. Traditional security measures often fall short in protecting data during model training, deployment, and inference, hindering the full potential of AI adoption.
Solution
Protopia AI addresses these challenges with its Stained Glass Transform (SGT) technology, which converts sensitive data and prompts into stochastic representations. This approach enhances data privacy and security throughout the AI lifecycle, improving accuracy and optimizing compute utilization. SGT enables organizations to securely transmit and process data across heterogeneous environments without exposing sensitive details, unlocking the ability to use the most performant GPUs in any environment, including on-prem, hybrid, or multi-tenant setups. The transformation process is computationally efficient, running significantly faster than cryptographic techniques, adding minimal latency to AI inference.
Target Audience
Protopia AI targets enterprise AI users and AI providers in sectors such as financial services, manufacturing, technology, and defense, who require secure data handling for AI model training, deployment, and inference.
Features
- Converts sensitive data into stochastic representations, maximizing distributional divergence between input vector embeddings and their transformed counterparts
- Retains semantic meaning, ensuring near-identical performance to plain-text representations in industry-standard LLM benchmarks
- Decouples ownership of plain-text data from the LLM implementation and infrastructure
- Supports any PyTorch module and Hugging Face Transformer for seamless integration into existing training loops
- Can be integrated into existing training loops with negligible impact on training throughput
- Allows enterprises to narrow the attack surface at the data layer
- Compatible with NVIDIA DGX Cloud for accelerated SGT creation
- Facilitates secure Retrieval-Augmented Generation (RAG) implementations