TVBS on AWS: How AI Is Changing Media Workflows

In an era where media consumption habits are rapidly evolving, TVBS has recognized that maintaining competitive advantage requires not just an upgrade in content strategy but also a significant enhancement in software development capabilities and foundational technology infrastructure. Over five years ago, TVBS established a new department led by Chen Hongyi, the Deputy Director of the Digital Development Center under the Strategy and Innovation Division. This team operates with a mindset similar to that of a software product company, focusing on technical selection, architectural design, and product validation with a core emphasis on rapid delivery and continuous iteration, continuously exploring new growth avenues for the company.

Embracing Agile Culture with IaC and Containerization

Discussing the team’s management and technical philosophy, Chen emphasizes that one of TVBS's core principles is to let technology lead the way. For the team, technology is not merely a passive support function but a proactive capability that must be established in advance to respond swiftly to market opportunities.

To turn the idea of "Trial & Error" into reality, TVBS abandoned the traditional procurement and setup model of on-premise data centers, opting for AWS as a vital cornerstone of their modern infrastructure. This shift not only involved migrating infrastructure but also fundamentally transformed the work logic and delivery cadence of the engineering team.

"Currently, we have migrated up to 95% of our workload to Amazon EKS," Chen states. For the team, the value of containerization lies not only in enhanced deployment efficiency but also in maintaining high consistency across local, testing, and production environments, significantly reducing the communication and debugging costs associated with environmental differences. Testing environments, which previously required manual assembly, can now be prepared in a much shorter time frame, allowing for resource scheduling in mere seconds. This enables the team to adopt a rapid iteration approach, validating and correcting products swiftly.

Optimizing the foundational architecture has also redefined the engineering team's working approach. Chen notes that developers used to aim for perfection in a single attempt; however, on a modern technological foundation, the team now emphasizes launching a product in the right direction quickly and then continuously optimizing through iterative processes.

Engineering Operational Excellence with SRE and DevOps

As the development phase accelerates, TVBS is simultaneously restructuring its operational model by forming a lean Site Reliability Engineering (SRE) team. This team integrates reliability, automation, and scalability into the daily engineering framework, making DevOps a practical working methodology rather than just a concept.

Reflecting on the inception of the SRE team, Chen outlines a clear objective: "To transition a high percentage of operational tasks from manual to automated processes, gradually achieving a state where systems can autonomously check and repair themselves without human intervention." Today, TVBS employs Infrastructure as Code (IaC) principles to manage its infrastructure programmatically, utilizing Terraform and AWS native tools to standardize and automate deployment, scaling, and recovery processes. When servers or services encounter issues, the system can trigger automatic recovery mechanisms, significantly reducing the frequency of manual interventions by engineers during off-hours.

For instance, when a server malfunctions, engineers previously had to manually investigate and resolve the issue. Now, supported by IaC and automated mechanisms, even if multiple servers face problems simultaneously, the system can initiate automatic recovery processes, greatly minimizing reliance on immediate human intervention. This foundational engineering capability directly enhances the service experience for TVBS’s external customers. To ensure that their digital services and websites perform reliably under high traffic and cross-regional transmission scenarios, the team has re-evaluated the overall data flow and transmission architecture, continuously optimizing for high availability.

"We have redesigned the entire data flow and transmission architecture, with the reassessment of CDN and deployment of Amazon CloudFront being just one aspect," Chen explains. By combining edge computing with precise scheduling of cloud resources, this modern architecture not only significantly reduces the misuse of IT resources but also effectively addresses cross-border transmission latency challenges, robustly supporting TVBS's extensive global digital traffic. Moreover, the SRE team acts as internal architectural consultants, assisting the development team in evaluating database, storage, and system design selections, thereby enhancing collaboration between front-end and operational teams and ensuring greater consistency in technical decision-making.

Innovating New Business Models: TVBS’s Lab Team and AI Product Development

Building on a solid infrastructure, another hallmark of TVBS's transformation is the internal "Lab Team." Chen positions the Lab Team as the company’s product experimentation unit, operating most similarly to a software startup. "We focus on developing digital products that are not directly related to traditional television programming, exploring how software innovation can uncover the next generation of growth opportunities for the company." This agile team, which comprises design, product, and engineering capabilities, can produce prototypes within a month with minimal communication friction and high collaboration density, swiftly validating whether new ideas have market potential.

To expedite the exploration of new business models, the Lab Team has aggressively integrated AI technologies. After evaluating numerous AI solutions available in the market, the team ultimately chose Amazon Bedrock as their core AI infrastructure.

"In the era of generative AI, betting on a single model poses significant risks," Chen points out. For the Lab Team, the key is not to chase after a single trending model but to maintain flexibility with multi-model support while balancing data security, management efficiency, and future scalability. Thus, the team selected Amazon Bedrock as the foundational platform for their AI applications. This enables them to seamlessly connect the most suitable top-tier models (such as the widely favored Claude family), eliminating cumbersome API connection and management barriers while ensuring enterprise-grade data security.

Today, the Lab Team's explorations are gradually reflected in their product outcomes. The team plans to launch a weight loss app that integrates AI functionalities, utilizing AI image recognition technology to accurately analyze calories through photographs and generate personalized dietary suggestions in real time. Daily, based on user conditions, it intelligently assigns task cards, transforming weight loss habits into a unique card game experience. This initiative not only incorporates popular AI topics but also translates AI into concrete product features, moving towards commercial viability.

From on-premise to the cloud, from manual operations to automation, and from traditional development processes to agile iterations, TVBS is continuously reshaping the public’s perception of media companies. For engineers, this environment is not just about maintaining existing systems but also about having the opportunity to participate in large-scale media traffic scenarios, modern infrastructure restructuring, and validating new products from concept to reality.

Chen recognizes the strength of Taiwanese engineers and hopes to showcase that even in an industry like television, which may seem traditional, software technology can be a core driver of industry evolution. "If you are eager to break the mold, want to validate the value of technology in real traffic scenarios, or aspire for your engineering skills to be recognized on a larger stage, we invite you to join us in writing the next chapter."