MLtwist Awarded $590K Contract for Multi-Modal Labeling

The U.S. Transportation Security Administration has awarded MLtwist a $590K contract for the labeling and processing of multimodal AI data.

The contract comes after completion of a successful aviation security pilot program in collaboration with Sandia National Laboratories in May 2024. Sandia used MLtwist’s technology to set up a reusable no-code AI data pipeline for labeling the DICOS data standard, which provides essential training for developing new object detection algorithms for security screening applications. The MLtwist-powered pipeline cut Sandia’s data processing time by 60% from eight to three weeks.

“New AI capabilities applied to data processing, working in conjunction with human intelligence, deliver higher data quality with faster turnaround times,” said David Smith, founder and CEO of MLtwist. “Thanks to the TSA’s Open Architecture initiative, which also supports the development of AI for software and hardware, we are excited to be one of the many pieces involved in the future of aviation security.”

The TSA Open Architecture initiative is focused on standards-based, interoperable software and hardware, which allows a wide range of industry partners to develop improved sub-components like detection algorithms, user interfaces, or reporting systems. The Digital Imaging and Communications in Security (DICOS) standard allows for significant improvements in the capturing and sharing of information by enabling security equipment’s images and their associated data to be captured, transmitted, stored in a DICOS-compliant medium that is viewable on any DICOS-compliant workstation or device. TSA efforts around Open Architecture are captured in their published roadmap[1].

The TSA will use the new data processed by MLtwist for the advanced development and implementation of new threat detection models. This data will be delivered through a robust and reusable framework, which leverages commercial off-the-shelf technology, and automates data cleaning, transformation, labeling, and packaging. MLtwist ingests the DICOS scans and preprocesses, transforms, and then delivers the resulting labeled data as a JSON or DICOS package. The process takes place in a secure environment, with data versioning and tracking provided at every step.

Boilerplate

MLtwist delivers secure, transparent, comprehensive, end-to-end AI data pipelines that automate the entire data preparation, automated labeling, human in the loop labeling, and management process. Supporting a wide array of data types, including video, audio, text, image, and even the 3D DICOS format common in security applications, MLtwist cuts the time and effort required for data processing. This level of automation not only speeds up model deployment but also allows data science teams to concentrate on enhancing data quality and developing models.

[1] https://www.tsa.gov/for-industry/open-architecture