Why DataMasque tips Amazon Nova Forge as a ‘game changer’ for custom AI

Harnessing more data, not just access to models, the key to custom AI and delivering business value.

AI is seen as a major differentiator, but if organisations are all accessing the same off-the-shelf models, there's little true differentiation, according to Grant de Leeuw, co-founder and CEO, DataMasque.

This is when custom AI models come into play, and at AWS Re:Invent 2025, the cloud giant revealed its Nova Forge technology.

Nova Forge blends proprietary data with Amazon Nova-curated datasets at every stage of model training.

“I think that’s going to be an absolute game changer, allowing organisations to start building their own models,” said de Leeuw.

This new capability can harness domain-specific “knowledge” without losing the AI model’s foundational reasoning capabilities.

“The value for organisations is the data that sits behind the firewall, and I think Nova Forge is going to allow them to start building specific models for their use cases trained on their own data,” he said.

It represents a decisive shift away from generic AI and RAG-only architectures toward fully trained, purpose-built models. It promises benefits like cost savings, accuracy and less hallucinations because it's trained on organisational data.

“Very few people in the world have the skill set to build a model, but Nova Forge is allowing them to start operationalising AI by leveraging their own data,” he said.

DataMasque specialises in supporting organisations to de-identify sensitive customer data using synthetically identical data and de Leeuw is excited about the possibilities of custom AI.

One application is building models that handle natural‑language instructions for data masking such as understanding how best to de‑identify data while not breaking application logic. The other is helping customers build on their own data for AI applications.

DataMasque recently signed a strategic collaboration agreement (SCA) with AWS with a view to collaborating across regulated markets such as financial services, healthcare, government and insurance.

With this in place, DataMasque will embed Nova Forge into its own masking workflows and enable customer organisations to use Nova Forge on de‑identified, high‑fidelity data.

“We can really unlock a lot of that data, particularly for banks where it's sitting in core systems, and make it available for training,” he told CRN Australia.

As partners focus on real-world AI applications, control over data, not just access to models, could be a vital ingredient in delivering business outcomes.

“The next word after AI is data — and we can provide the data that ensures AI models never get the real customer information, but still work as a true representation of production,” he ended.

Rosalyn Page travelled to AWS Re:Invent 2025 as a guest of AWS.