In 2020, Gamania’s big data team laid out a unified architecture of data collection using Dataflow and Dataproc for all Gamania’s subsidiaries and began to store all its business data in Cloud Storage. Based on the data, the team quickly built a centralized data lake in BigQuery to enable speedy data retrieval and big data analytics. Chasel Su, Head of Big Data Center and Co-Head of AI Innovation Lab at Gamania, notes that with key technologies like column-oriented storage, serverless autoscaling, distributed nodes, query optimization, BigQuery not only supports 1.8x faster query speed than other similar offerings, but also allows the quick setup of Gamania’s centralized data warehouse through seamless integration with datasets in Cloud Storage and data sources like Google Ads and Google Analytics that Gamania’s subsidiaries rely heavily on.

Prior to the adoption of Google Cloud, whenever Gamania needed to conduct cross-subsidiary data analytics, its data scientists needed more than one week to gather and process data from databases deployed in various environments. Now, even Gamania’s non-engineer employees can use BigQuery to run the analytics they need on their own and retrieve reports within a few minutes. 

This way, Gamania is able to execute effective and timely data-driven decision-making. For example, when its operations team runs campaigns, it needs to adjust strategies based on consumers’ engagement. Before using BigQuery, the team had to wait for two days to receive data for campaign performance evaluation. The bigger a campaign was, the longer the team had to wait for the data.

Now, by leveraging Dataflow to stream data in real time to BigQuery for analytics, the Gamania team can monitor how consumers respond to their campaigns instantly and swiftly change their strategies when needed.

Having consolidated its centralized data storage and analytics, Gamania decided to maximize the value of its data through the machine learning (ML) technology. For its ecommerce and online media business, it developed several AI recommendation systems by leveraging Dataproc to process data and tag products automatically, and GPU resources in Compute Engine to train ML models. To provide the most relevant content, Gamania also employs Pub/Sub and Dataflow to support real-time data streaming and ensure that its recommendation system can reflect users’ latest behavior within three seconds. 

Since its recommendation system is trained with a large number of data and capable of generating recommended content based on user behavior under different circumstances like seasons and time, Gamania is able to provide hyperpersonalized experiences and has improved the conversion rate on its ecommerce sites by up to 17x, as well as the interaction rate on social media by up to 300 percent.

Besides the AI recommendation system, Gamania has developed more than a dozen AI solutions deployed in GKE for internal use, including a customer service chatbot named Vyin Brain that it launched in two of its most popular games and reached a resolution rate of 97 percent. 

One of Vyin Brain’s major features is the ability to detect user context, which is supported by a number of complex analyses like intention, emotion, and entity relationship analysis. To facilitate the development process, Gamania needed to first merge all the analysis models and resolve the integration issues. To that end, the team used the pretrained models and templates in Vertex AI to quickly build a prototype and identify the potential problems, which made it much easier to optimize and finetune its self-developed models afterwards.

With its strong expertise in AI innovation, Gamania will start providing AI solutions to enterprises in various sectors across Taiwan, Southeast Asia, and North America from the fourth quarter of 2024 through its newly established brand Vyin AI. At the same time, the corporate group plans to expand its use of the Gemini models to further increase its employees’ productivity, such as image generation and workflow optimization.