Akatsuki AI Technologies supports TTDC’s VLM validation and PoC development using a digital twin

prtimes.jp reported on September 29, 2026, that Akatsuki AI Technologies is supporting TTDC’s VLM validation and PoC development using a digital twin. We have not been able to verify the information from a primary source; please refer to the original for details.

TTDC’s work involves using a VLM on a digital twin and implementing HMI generation, a cloud operating environment and connections to other systems. Because the validation covers surrounding implementation as well as the model itself, it highlights the importance for companies using AI in manufacturing and design of having a structure that allows requirements definition and development to proceed in parallel.

What is new in this announcement is that Akatsuki AI Technologies is emphasizing a support structure combining dispatched engineers and a technical lead, rather than one-off contract development. The work addresses VLMs, HMI, a cloud operating environment and connections to other systems at the same time, so the scope of validation covers implementation and operating conditions, not just the algorithm. At the same time, this extends an established effort to use VLMs on digital twins; rather than a splashy feature announcement, it appears to be a way of organizing the process for turning technical validation into a PoC. Our previous report, Akatsuki AI Technologies deploys humanoid and quadruped robots at Oarai Ocean Fireworks Festival 2026, covered the implementation of robots at an event attended by the general public. This time, the focus is not on an event deployment but on validating a VLM on TTDC’s digital twin. Within the same physical AI context, the former focuses on public-facing deployment, while the latter focuses on the depth of design and validation. In other words, the company is pursuing public-facing deployment and internal validation in parallel. The broader industry implication is that the value of VLMs depends not only on model performance but also on cloud execution environments, HMI, connections to external systems and iterative work on requirements definition. Since the work uses a digital twin, the extent to which assumptions on the operational data and simulation side are in place will affect the PoC’s success. Akatsuki AI’s emphasis on “technical support as an organization” suggests that the work requires not only implementation but also a process for feeding issues back to the team considering the project. Unresolved questions include the PoC’s scope, evaluation metrics, implementation targets and whether the effort will advance to production use.

When Japanese companies validate combinations of digital twins and VLMs, the challenge is not just model accuracy but also how to integrate HMI, cloud environments and connections to existing systems. In automotive-related design and validation, companies with structures that can handle requirements planning and implementation in parallel, as TTDC does, are likely to have an advantage.