Enables high-performance AI agent that can be activated by one tap with Gemini
Empowers quick development and deployment of video generation AI agents with Agent Development Kit
Provides highly accessible, localized multimodal AI features to empower users in emerging markets
With Google AI, Transsion is building next-generation AI user experiences on mobile and digital devices for all kinds of users.
With Google AI, Transsion is building next-generation AI user experiences on mobile and digital devices for all kinds of users.
Building a next-generation AI ecosystem
As a leader of smart end devices and digital services in emerging markets, Transsion operates in more than 70 countries and regions across Africa, Southeast Asia, South Asia, Middle East and Latin America, owning several well-known mobile phone brands. According to IDC, Transsion’s worldwide smartphone market share was 8.7% in 2024, ranking fourth.
Since its foundation, Transsion has been committed to providing cutting-edge technology to users around the world, continuously developing easy-to-use and innovative AI features for its digital devices. As the AI technology entered the third wave of evolution, Trassion’s development work has a new goal—to create a next-generation AI intelligent ecosystem that can think proactively and execute tasks automatically.
“Transsion’s vision for AI phones is to proactively think like a human, execute like a machine, and solve tasks end-to-end,” explains Tuanwei Shi, general manager of Transsion’s TEX AI Center. “In the past, users had ideas and needed tools to realize them. In the future, AI will proactively think for them, even completing tasks without user intervention.”
Google Cloud offers diverse, ever-advancing AI models that meet our technical needs, and multilingual capabilities that can greatly simplify product development for a diverse audience.
Tuanwei Shi
General Manager, TEX AI Center, Transsion
To bring this vision to hundreds of millions of users in emerging markets, with diverse languages and cultures, Transsion needed a powerful AI platform with cutting-edge modal capabilities, multimodal interaction support, the ability to build complex agents, as well as wide global reach. The team turned to Google Cloud, because its AI models fulfill these requirements.
“We not only aim to build top-notch AI features, but also to make them accessible for all,” notes Shi. “Google Cloud offers diverse, ever-advancing AI models that meet our technical needs, and multilingual capabilities that can greatly simplify product development for a diverse audience.”
Realizing comprehensive multimodal AI features across digital devices with Google AI
Leveraging different models of Gemini, coupled with new technologies such as Agents2Agents (A2A) and Model Context Protocol (MCP), Transsion developed Ella, a voice assistant, and several AI agent products that support diverse interactions from text to real-time audio or video. Powered by the agent capabilities of Gemini 1.5 Flash, Ella understands user instructions better and responds instantly, requiring only 0.23 second to receive the first token.
Based on the Ella model, Transsion has successfully developed a series of product features that help facilitate users’ daily lives, including one-touch screen inquiry, smart screen recognition, AI search, call summaries, and AI writing.
On top of that, the Transsion team built an AI image and video generator by merging the text refinement capabilities of Gemini with Imagen and Veo. Requiring only a few keywords, this feature refines users’ prompts and allows Ella to quickly generate quality images and videos in the style desired.
To further lower the barrier to video creation for users in emerging markets, Transsion has rapidly developed and deployed a sub-agent for image-based video generation using Ella’s lightweight multimodal generation tool and Agent Development Kit (ADK). In this architecture, Transsion uses large language models like Gemini for decision-making, reasoning and planning, and leverages Cloud SQL or Gemini Enterprise Agent Platform Agent Engine Sessions to store context and status during individual tasks, as well as Firestore or Gemini Enterprise Agent Platform Vector Search to store knowledge that requires long-term retention or information available for semantic search, supporting the sub-agent’s long- and short-term memory.
“Our AI product development has greatly benefited from the AI models of Google Cloud, such as Gemini, Imagen, and Veo, because of their high performance and interoperability,” says Shi. “The abundant AI development tools and infrastructure resources of Google Cloud also facilitate the iterations of our multimodal features moving forward.”
Our AI product development has greatly benefited from the AI models of Google Cloud such as Gemini, Imagen, and Veo, because of their high performance and interoperability. The abundant AI development tools and infrastructure resources of Google Cloud also facilitate the iterations of our multimodal features moving forward.
Tuanwei Shi
General Manager, TEX AI Center, Transsion
Empowering users in emerging markets while expanding AI ecosystem
With its significant presence in emerging markets, it is important for Transsion that technology advancements address users’ real needs. For example, for children who lack help for their homework, Transsion built an image-based question-and-answer AI model powered by the multimodal and knowledge base capabilities of Google Cloud. Trained for education scenarios, Ella can accurately answer difficult questions across various subjects. Users only need to take a picture of a question with their phone.
In terms of data security, Transsion plans to establish a unified and trusted identity authentication and data backflow system for the multi-cloud AI agents that support Ella, strictly adhering to data sovereignty and regulatory requirements. In markets where data sovereignty is required, it will deploy independent regional centers to receive, process and store all agent-backflowed data within each region. Additionally, Transsion will build a service identity authentication center according to the SPIFFE/SPIRE standard. Each AI agent, upon startup, will use its native identity from its cloud platform, such as Google Cloud Service Account, to authenticate with the SPIRE server.
Through the in-depth collaboration with Google Cloud, we’re building a next-generation agent and ecosystem engine tailored for emerging markets, ensuring that cutting-edge technology benefits every user.
Tuanwei Shi
General Manager, TEX AI Center, Transsion
Following identity verification, the SPIRE server will issue a SVID, an automatically rotating, cloud-platform-independent identity credential, for two-way TLS encrypted authentication of all communication between the agent and other components. This enables unified management of agent identities in a multi-cloud environment and significantly enhances security.
Next, Transsion plans to expand its AI ecosystem by deepening its collaboration with Google Cloud and inviting developers worldwide to jointly build an open ecosystem that allows cross-border knowledge sharing. It will continue using the AI models of Google Cloud to develop more localized features and enhance Ella’s generation capabilities and interaction efficiency.
“Through the in-depth collaboration with Google Cloud, we’re building a next-generation agent and ecosystem engine tailored for emerging markets, ensuring that cutting-edge technology benefits every user,” notes Shi.
Transsion is a leading player in the mobile industries of global emerging markets, holding a significant global market share.
Industry: Manufacturing
Location: China
Products: Google Cloud, Agent Development Kit, Cloud SQL, Firestore, Gemini, Imagen, Veo, Gemini Enterprise Agent Platform (formerly Vertex AI)
