Digital business and technology services provider NTT Data and data specialist Bifrost AI today announced the results of their collaboration, leveraging synthetic data to develop AI models. The companies target use cases for federal agencies and decision-makers seeking lower-cost, higher-efficiency roads to AI productivity gains.
Collaboration set out to determine whether synthetic data would decrease AI model development costs
Federal agencies increasingly rely on AI to monitor critical environments, from coastal waters to disaster zones. But obtaining the real-world satellite imagery needed to train these systems is expensive, slow, and sometimes impossible due to restricted locations or infrequent events.
Over the past year, NTT DATA’s Innovation Center collaborated with Bifrost AI to evaluate whether synthetic data (computer-generated scenarios designed to replicate real-world conditions based on existing data) could accelerate AI model development while maintaining or improving quality and reducing costs.
The pilot was designed to answer a fundamental question in satellite analytics: whether synthetic data could meaningfully improve the economics and timeline of AI model development.
Key results: 300x iteration speed and major cost reductions
To the companies involved, the pilot program was overwhelmingly successful. Synthetic data generation enabled development teams to produce on-demand training datasets for rare scenarios, rather than waiting months for satellite tasking and real-world data collection.
This approach demonstrated up to 300x faster iteration speeds in specific test cases, allowing teams to refine models and pursue new customer requirements within days rather than quarters.
“We’ve seen the core validation that synthetic data works at scale across our defense and maritime customers. What really stood out here was the speed differential: in specific test cases, hitting 300x faster iteration exceeded even our own internal benchmarks,” Bifrost AI co-founder and CEO Charles Wong told Channel Insider.
Synthetic data also eliminated many traditional cost drivers associated with satellite AI development, including satellite tasking, restricted location access, and manual image annotation. Since synthetic images are generated with accurate labels automatically, training cycles that previously required months and substantial capital investment could be completed in significantly shorter timeframes.
How synthetic data improves planning for federal and maritime agencies
Early estimates suggest potential reductions in data acquisition costs of up to 70 percent in scenarios where synthetic data supplements real-world training sets.
“Our work with NTT DATA proved that the hybrid approach (synthetic + human data) works at enterprise scale for satellite intelligence. We’re already rolling out these capabilities with the U.S. government and defense primes, such as Anduril, the U.S. Air Force, and others,” Wong said.
The synthetic data approach demonstrated in this pilot offers several advantages for these teams:
- Rapid prototyping of new detection capabilities
- Greater agility in responding to safety-critical operational requirements
- More predictable cost structures, even for high-risk or rare scenarios
- Reduced dependency on limited or sensitive satellite tasking resources
“The true potential of AI is not in language learning models or image generation, it’s in the systems that will transform human life by taking over what’s dangerous and critical in the real world,” Wong said. “And synthetic data is what will get us there.”
Why service providers need innovation partners for future AI capabilities
NTT Data offers a range of technology and business services to clients across industries, including banking, healthcare, life sciences, and public-sector agencies. By working with Bifrost, the provider is expanding its expertise in delivering AI-enabled solutions to customers.
For providers like NTT Data, partnerships with synthetic data specialists are increasingly essential. Federal customers now expect AI partners to deliver proven, low-risk pathways to autonomous system development.
“Our customers face mission-critical scenarios every day. When a ship or a satellite can’t function, that’s not inconvenient. It can’t happen,” Wong said. “By enabling faster, safer training at the development level, we’re building the infrastructure that determines who can deploy autonomous systems with confidence. That’s where the real returns are made.”
Earlier this year, we covered NTT Data’s collaboration with Corvic AI, another example of the way the services provider is growing its AI capabilities and offerings.