In the rapidly evolving landscape of digital transformation, Anand Naidu stands as a pivotal figure in bridging the gap between traditional software engineering and the new frontier of AI-native development. With an extensive background in both frontend and backend systems, he has spent years navigating the intricacies of coding languages and the architectural demands of global tech ecosystems. As the Chief Product Officer at Shopee and a leader within Sea Limited, Naidu has spearheaded the rollout of advanced AI tools like Codex, transforming how massive engineering teams handle the weight of systemic complexity. This discussion explores the shift from manual coding to agentic workflows, the unique role of Southeast Asia as an innovation hub, and the cultural redesign required to thrive in a world where developers act as system orchestrators.
With weekly active usage rates reaching nearly 90% in large engineering organizations, how does deep contextual awareness change how developers handle complex microservices? What specific metrics should leaders track to measure the shift from localized productivity to structural effectiveness?
The true friction in a massive microservices architecture isn’t the act of typing syntax; it’s the exhausting detective work of tracing dependencies across a sprawling, often undocumented web of services. When you have an 87% weekly active usage rate for a tool like Codex, you start to see a visceral change in the office atmosphere—the quiet frustration of “navigating the maze” is replaced by a more fluid, creative energy. Instead of spending hours squinting at legacy logic written by someone who left the company three years ago, developers use deep contextual awareness to instantly grasp how a change in one service ripples through the entire ecosystem. Leaders should move away from tracking simple lines of code and instead monitor the “time-to-context”—how quickly an engineer can contribute to an unfamiliar service—and the reduction in reliability incidents under peak loads. It is about shifting the cognitive load away from the “how” of the plumbing and toward the “why” of the architecture, allowing teams to maintain a high-level perspective even when buried in millions of lines of code.
Integrating AI agents directly into CI/CD pipelines allows for autonomous reasoning and edge-case detection. How do these agents systematically help pay down technical debt, and what steps are required to ensure they maintain engineering discipline?
The transition from passive autocomplete to an integrated agentic workflow is a fundamental leap, where the AI becomes an active participant in the CI/CD pipeline rather than just a helpful ghost in the editor. We’ve seen that among our developers who rate these tools highly, about 73% would recommend them to colleagues, largely because these agents take on the “grunt work” of generating exhaustive test coverage and surfacing those rare, nightmare edge cases in distributed systems. To maintain discipline, we implement a workflow where the agent first reasons through the product requirements, proposes a test-driven implementation, and then allows the human to audit the logic before it ever hits production. This systematic approach ensures that we aren’t just shipping features faster, but we are actually cleaning up the underlying “mess” by allowing the AI to prototype alternative, cleaner implementations that a time-pressed human might overlook. It turns the struggle of paying down technical debt into a continuous, automated process that happens every time a developer pushes a single line of code.
As implementation tasks are abstracted away, the developer’s role is evolving into that of a system orchestrator. How do you prepare a workforce for this shift toward architectural design, and what are the trade-offs when product judgment replaces manual coding?
Preparing a workforce for this shift requires a complete reimagining of what it means to be a “senior” engineer, moving the focus from mastery of a specific language to mastery of system design and product intent. We are seeing the developer role evolve into a “system orchestrator,” where the primary skill is no longer wrestling with compilers but orchestrating complex, AI-driven workflows and making high-stakes decisions about system resilience. This transition is not without its trade-offs; as we rely more on AI for implementation, there is a risk of losing the “muscle memory” of manual coding, which is why product judgment becomes the new North Star. We encourage our teams to spend the bulk of their time on tasks that require human empathy and strategic foresight, ensuring that the speed gained from AI doesn’t lead us into a “velocity trap” where we build the wrong things faster. It is a cultural redesign where the reward is no longer a bug-free function, but a robust, scalable system that solves a real-world problem for millions of users.
Southeast Asia has a history of leapfrogging tech cycles, such as moving directly to mobile-first ecosystems. How does solving for fragmented, multilingual markets make the region a unique proving ground for AI-native development, and what can other hubs learn?
Southeast Asia is a beautifully chaotic mosaic of languages, payment systems, and logistics networks, which makes it the most demanding, and therefore the most rewarding, laboratory for AI-native development. Because our builders have always had to solve for fragmentation—like creating super-apps that work across thousands of islands with varying internet speeds—they have a natural grit and a “leapfrog” mindset. Other global hubs can learn from the way our local community uses AI agents to bridge these gaps, building systems that are inherently flexible and capable of handling hyper-localized complexity without bloating the codebase. We are moving directly into an AI-first era because we don’t have the luxury of sticking to legacy desktop-centric processes; our developers are already comfortable with the idea that the “interface” is whatever the user needs it to be at that moment. The region’s success comes from this necessity to innovate under pressure, turning linguistic and logistical hurdles into the very features that define the next generation of global software.
Regional hackathons are often used to democratize access to advanced AI primitives for local developers. How do these initiatives accelerate the move from raw curiosity to deploying scalable applications, and what defines a compounding talent ecosystem?
When we take a Codex Hackathon Series to cities like Singapore, Jakarta, Taipei, and Ho Chi Minh City, we aren’t just hosting a competition; we are handing the world’s most advanced building blocks to a hungry generation of creators. These events are crucial because they lower the barrier to entry so drastically that a developer can move from a “what if” curiosity to a fully functioning, scalable application in just a few hours. A compounding talent ecosystem is defined by this speed of iteration—when you upskill a developer in Vietnam or Taiwan today, they immediately begin teaching their peers, creating a ripple effect of technical literacy that accelerates the whole region. We’ve seen incredible success stories where teams built complex, AI-native tools during these events that would have previously taken a well-funded startup months to prototype. This democratization of access ensures that the next great AI innovation is just as likely to come from a small coffee shop in Indonesia as it is from a massive lab in Silicon Valley.
What is your forecast for AI-native software development?
I foresee a total reconfiguring of the engineering hierarchy where the “cost of execution” drops so low that the only remaining bottleneck is the quality of our ideas and our architectural foresight. Software teams will become significantly more leveraged, with a single orchestrator directing a fleet of specialized AI agents to build, test, and maintain systems that are far more resilient than anything we could code by hand today. We will see the “developer” title eventually fade, replaced by “systems architects” who focus on the ethical, social, and structural implications of the software they guide into existence. For those who embrace this human-AI collaboration now, the future holds a world where we spend less time debugging the past and more time building a future that was previously considered too complex to even imagine.
