Deloitte has launched an Open Model Engineering practice in India and other key global markets to help enterprises build, deploy, operate and scale agentic AI applications using a combination of open and proprietary AI models.
The new practice will initially operate across North America, Europe and Asia Pacific, including India. It will focus on developing enterprise AI applications and products using open models and full-stack open-source AI frameworks, while allowing organisations to choose models and deployment environments based on their business and technology requirements.
Deloitte said the practice is designed to give enterprises greater flexibility in AI deployment, along with more predictable costs, control over data and intellectual property, and greater choice over where AI models are deployed.
Sathish Gopalaiah, President, Consulting, Deloitte South Asia, said enterprises are increasingly moving from a model-first approach toward an architecture-first approach. According to Gopalaiah, open models allow organisations to shape AI around their own data, processes and business context rather than changing their operations to fit the limitations of a particular model.
The practice will help clients evaluate and select between proprietary models, open models, agentic platforms, cloud services and on-premises infrastructure. Deloitte will also support organisations in managing token economics by selecting suitable models, infrastructure and deployment approaches for different workloads.
Another area of the practice will be fine-tuning open models for geographical, linguistic and cultural requirements. Deloitte said this can give organisations greater control over where their models run, what data is uploaded and how intellectual property and the uniqueness of generated outputs are protected.
The practice will initially focus on enterprise AI applications built using NVIDIA’s Nemotron family of open models and NIM microservices. Deloitte will also help clients develop Sovereign AI stacks using open-source AI frameworks, fine-tune open models and strengthen cyber defences with open harnesses.
Kari Ann Briski, Vice President of Generative AI at NVIDIA, said, “Enterprises need both open and proprietary models for different workloads. Deloitte’s Open Model Engineering practice, built on NVIDIA Nemotron models, gives developers greater control over enterprise data and proprietary information while enabling secure, responsible deployment of Agentic AI.”
Deloitte will also use its Zora AI digital workforce platform as part of the new practice. The platform includes agentic harness capabilities built on NVIDIA Nemotron open models and NIM. According to Deloitte, this will provide clients with a way to deploy, continuously optimise and govern AI agents using open models within their own environments.
The practice is also expected to expand Deloitte’s engineering capabilities through investment in specialised talent. The company plans to hire, train and certify forward-deployed engineers globally through fiscal year 2027. These engineers will work directly with Deloitte clients to implement open-model solutions and will form the operational foundation of the new practice.
Deloitte said enterprises are currently considering several factors when deploying AI, including the ability to select the appropriate model and architecture for different workloads, manage costs as agentic AI usage grows, determine where and how models are deployed, and maintain control over enterprise data, intellectual property and model behaviour.
The launch expands Deloitte’s AI services portfolio as enterprises explore different approaches to building and deploying AI systems. The Open Model Engineering practice is intended to support organisations in combining open and proprietary models while developing AI applications suited to their individual business, regulatory and technology requirements.
