COMPUTING INFRASTRUCTURE

AI computing infrastructure

The foundation for intelligence that evolves.

MBTIX is building its own data center for AI research and our AI services. Planned workloads include preparing business, customer-interaction, media and driving data, developing and evaluating models, and running inference in our products.

Research batch processing and online inference have different capacity and response-time requirements. The facility is being designed around workload, performance and operating cost to support both research and business operations.

A section model showing optical cabling, server racks and underfloor cooling in a planned data center
03 / INFRASTRUCTUREHow we plan to use our compute facility
  1. Operating and support records are organized for the algorithm and evaluation task at hand.

  2. Dispatch and dialogue models can be compared under consistent conditions against operational needs.

  3. The facility under development is designed around workload and operating cost to support research and our own services.

Conceptual view of the infrastructure

Capabilities and value

Our data center is under construction to support AI research and our AI services.

We are planning computing resources for data processing, model development and evaluation, and product inference across our businesses.

Data-processing environment

Develop computing support for organizing, analyzing and evaluating operational and real-road data.

Algorithm and model evaluation

Support development and evaluation across dispatch, routes, dialogue and operational suitability.

Connections to our AI services

Translate product use cases and workloads into infrastructure requirements.

Performance and resource planning

Consider performance, workloads, resource costs and management for each intended use.

Where it applies

Research and development

Provide a foundation for continued data work, algorithm development and model validation.

Product operations

Plan around AI-enabled service requirements and future product development.

Research–operations feedback

Use service needs and research questions to improve the computing environment.

Preparing operating records for model evaluation

For dispatch evaluation, the planned workflow pairs booking conditions with actual assignments and compares candidate matches under consistent conditions.

Separating batch work from time-sensitive requests informs facility design and future product improvements.

Define requirements

Review requirements from products, research and target data.

Plan the environment

Plan for processing, evaluation and operational workloads.

Develop the facility

Develop the facility and establish the conditions required by research and services.

Improve for actual use

Use performance, workload and operational findings in future development.

Value we create with partners

Our research and service requirements guide the prioritization of data processing, model evaluation and inference, and the allocation of computing resources.

The facility is being developed as a shared foundation for research and product improvement across mobility, digital media and AI services.

CONTACT MBTIX

Talk to us about business and technology

Contact us about product adoption, business collaboration or technology. We will direct your inquiry to the appropriate team.

Contact us