How MLtwist Streamlined AI Data Processing for Sandia National Laboratories

The Use Case

Sandia National Laboratories is helping to advance aviation security by integrating AI into threat detection systems. However, training these AI models requires vast amounts of accurately labeled, multimodal data—especially in specialized formats like DICOS (Digital Imaging and Communications in Security), which is one major file standard used to encode 3D scan data.

THE CHALLENGE

“Managing Multi-Modal Labeling While Ensuring Speed, Scalability, and Compliance”

Sandia faced key challenges:

• Time-Intensive Data Processing: Traditional methods took up to eight weeks to prepare DICOS scan data for AI training.
• Complex Multi-Modal Labeling: Security screening data includes a mix of images, metadata, and 3D scan formats, requiring precise annotation.

Scalability & Compliance: The agency needed a secure, standardized pipeline that could support Open Architecture initiatives for interoperability.

MLtwist’s Solution: A Fully Automated AI Data Pipeline

• MLtwist delivered a fully automated AI data pipeline, optimizing the entire process from ingestion to final structured output:

  • Automated Data Cleaning & Transformation: MLtwist ingested raw DICOS scans, structured the data, and prepared it for annotation.
  • High-Precision Multi-Modal Labeling: Advanced AI-assisted workflows ensured accurate annotations across image, metadata, and 3D scan formats.
  • Seamless JSON Packaging and Delivery: Once labeled, the data was automatically formatted into JSON outputs, making it instantly usable for AI detection models.

  End-to-End Data Tracking & Security: Every step in the pipeline was versioned and fully traceable, ensuring compliance with strict security standards.

•  

Impact & Benefits

• • 60% Faster Processing: MLtwist reduced Sandia’s data preparation time from eight weeks to three weeks.
  • Improved AI Model Training: Consistently structured, high-quality labeled data accelerated the development of new threat detection models.

  Reusable & Scalable Framework: The pipeline provided a repeatable process for future AI-driven security initiatives.

•  

The Takeaway

Following a successful pilot with Sandia National Laboratories, MLtwist secured a $590K contract to process multi-modal AI training data for next-generation aviation security. By delivering a secure, automated, and scalable data pipeline, MLtwist is helping to enhance AI-powered threat detection capacities while reducing operational bottlenecks.