Skip to content Go to main navigation Go to language selector
Saab Global

Shaping the Future: Devaki Raj, Saab Strategy Office’s New Chief Digital and AI Officer, Discusses Artificial Intelligence with U.S. Senate Committee on Homeland Security & Governmental Affairs

3 min read

Devaki Raj, chief digital and AI officer in Saab, Inc.’s strategy office, was invited to participate in the U.S. Senate Committee on Homeland Security and Governmental Affairs’ hearing, “Governing Artificial Intelligence (AI) Through Acquisition and Procurement.” Recognized as one of the top minds in this space, Devaki’s September 14 testimony focused specifically on government AI contracts – the challenges companies face and how different procurement processes can lead to more long-term success for government projects.

Raj is the founder of CrowdAI, which was recently acquired by Saab, Inc. CrowdAI is a no-code computer vision application that supports both DoD and commercial customers through mission-specific AI. An expert in Machine Learning technologies, Raj has collaborated with the California Air National Guard to automate wildfire mapping using MQ-9 drones and with the U.S. government to counter narcotics trafficking in South America.

Devaki Raj, chief digital and AI officer in Saab Inc.’s strategy office

“Many of these (Artificial Intelligence) systems are not developed by the government but rather by the private sector,” said Sen. Gary Peters, chairman of Senate Homeland Security and Governmental Affairs Committee at today’s hearing. “This collaboration between the public and the private sector is crucial. It ensures that government is using the most effective AI systems. American companies are breaking new ground with these technologies, and we have the chance to share in the benefits of that incredible innovation.”

In her testimony, Raj explored the necessity of effective AI procurement vehicles through her experience collaborating with U.S. government entities. She identified four key points of focus to ensure effective government AI procurement:

  1. Commercial and off-the-shelf AI solutions need government-curated data to be mission-ready.
  2. AI procurement should include ongoing AI model training and the infrastructure to support that training.
  3. The rapid growth in open-source AI technologies necessitates rigorous testing and evaluation before and after procurement.
  4. Paths to programs of record for small businesses are necessary through project transition milestones. 

In her oral and written testimony, Raj emphasized that AI must be thought of as a journey, not a destination.

“Often, today, we see solicitations ask explicitly or implicitly for ready-to-deploy automated solutions; however, government missions are inherently unique. Mission-specific data are required for each algorithm because each problem posed by homeland security, intelligence, or public health is inherently unique,” wrote Raj.

She emphasized that successful AI modeling is not an out-of-the-box, one-time solution, but rather an ongoing project that requires continuous training and model adjustments using mission-specific data for ideal results.

“For operationally-ready AI, there is no functional difference between development and operations, as they are continuously interlinked. Therefore, we must move past AI being funded as a one-off software purchase and build procurement vehicles that bake in ongoing updates or service-level agreements. This is not only because AI needs re-training but also to provide the procuring officer with technology that is cutting edge.”

Leveraging her experience, Raj also emphasized the need to establish rigorous AI testing based on qualitative and quantitative metrics.

“On account of how procurement works today, companies often project future potential capabilities, regardless of if they are possible, to win federal awards. The risk of development failure comes later and falls on the government…Codifying government-wide standards for AI testing and evaluation would help mitigate unverifiable corporate claims.”

“In all phases of an AI project lifecycle, remember that machine learning, just like human learning, is a journey and not a destination,” Raj added.