KK Blog

菜单

Administrator
Published on 2026-05-17 / 2 Visits
0
0

"Sea Limited AI-Native Software Development: How Asia's E-Commerce Giant Deploys Codex Across Engineering Teams"

#Southeast Asia#Shopee#Codex#Sea Limited#AI-Native Engineering#Developer Productivity#Enterprise AI#AI Coding#OpenAI

Most companies deploying AI coding tools follow the same script: buy licenses, distribute them to engineers, wait for adoption metrics. Sea Limited is doing something different. The Singapore-based technology company, whose businesses span e-commerce (Shopee), digital entertainment (Garena), and digital financial services (Monee), is not merely rolling out Codex to its developers. It is building an organizational infrastructure around AI-assisted software development, from internal workflow standardization to a regional hackathon series that will carry the practice to developer communities across Asia Pacific.

The difference matters. The first approach treats AI coding tools as a product you deploy. The second treats them as a capability you construct. In the emerging landscape of AI-native software development, that distinction is beginning to separate companies that extract marginal productivity gains from those that restructure how software gets built.

The Scale of the Bet

Sea Limited is one of OpenAI's largest Codex customers in Asia, according to Thibault Sottiaux, Head of Codex at OpenAI. The company has rolled Codex out across its entire developer organization. Internal data shows 87% of users are weekly active users, with 73% recommending the product. These are adoption numbers that most enterprise software deployments never reach.

On May 15, 2026, Sea and OpenAI jointly announced the Sea x OpenAI Codex Hackathon, a regional series beginning in Singapore on June 6 and expanding to Indonesia, Taiwan, and Vietnam. The Singapore kickoff alone is expected to bring together more than 150 developers working on over 40 projects. Winning teams will receive up to US$30,000 in OpenAI API credits, with additional prizes including ChatGPT Pro subscriptions.

The hackathon is not a marketing event. It is a structural investment in an ecosystem. Sea is treating AI-native development not just as an internal efficiency play but as a regional capability that needs to be cultivated, practiced, and expanded.

David Chen's Framework: Structural Multiplier, Not Marginal Gain

David Chen, Co-Founder of Sea and Chief Product Officer at Shopee, laid out the strategic thinking in an interview published on OpenAI's blog as part of its Executive Function series. Chen's framing is worth examining closely because it represents a distinct philosophy about what AI coding tools are for.

"At Sea's scale, engineering isn't just about writing code," Chen said. "It's about managing large-scale systemic complexity across fragmented, hyper-localized markets. We see the ongoing developments in AI leading to a fundamental shift in how software is created and how our engineering teams operate at scale."

The key phrase is "structural multiplier." Chen does not describe Codex as a tool that helps engineers write code faster. He describes it as something that amplifies the engineering organization's ability to handle complexity at a scale that individual humans cannot manage alone. This is a different mental model from the one most companies use when they evaluate AI coding tools. The default frame is productivity: can engineers ship more features in less time? Chen's frame is capability: can the organization manage complexity that would otherwise be unmanageable?

Shopee operates across Southeast Asia and Taiwan, with a growing presence in Brazil. Each market has distinct payment systems, logistics networks, regulatory requirements, and consumer behaviors. The engineering challenge is not building one product, but building dozens of localized variants of that product while maintaining a coherent technical platform. AI agents that understand large codebases, can navigate cross-cutting concerns, and can execute operational tasks autonomously address this specific challenge in a way that autocomplete-style tools do not.

The Infrastructure Build: Beyond Tool Deployment

Sea's approach reflects the builder's mindset: the moat is not in the tool itself but in how you adapt, compose, and extend it. Three structural investments distinguish Sea's deployment from a standard license rollout.

Workflow integration at the organizational level. Codex is embedded into Sea's existing development workflows, not layered on top. This means integration with code review processes, CI/CD pipelines, and the internal tooling that Sea's engineering teams already use. The difference between "engineers use Codex on their own machines" and "Codex is wired into how the organization ships software" is the difference between individual productivity and organizational capability.

The AI Centre of Excellence. In April 2026, Sea established an Artificial Intelligence Centre of Excellence in Singapore, with support from Digital Industry Singapore (DISG). The centre is expected to create at least 100 R&D and innovation-focused roles over three years, spanning AI research, engineering, and product development. It focuses on three areas: foundational AI (model development and evaluation), scalable deployment (translating R&D into production systems), and AI-native talent development.

The Centre also deepens the application of Sea's in-house AI models, including Compass Max v3.5, a 245-billion-parameter large language model tailored for Southeast Asian languages and e-commerce contexts. Compass Max v3.5 and its variants already power AI features across Shopee, operating at a fraction of the cost of commercial LLMs. This means Sea is not just consuming external AI tools, it is building proprietary AI capabilities that complement them.

Ecosystem expansion through the hackathon series. The regional hackathon series serves multiple purposes. It builds brand awareness for AI-native development practices in markets where Sea operates. It creates a pipeline of developers who are familiar with Codex and AI-assisted workflows. And it generates practical use cases and feedback that inform how Sea and OpenAI refine their tools for the Asia Pacific context.

The Broader Context: Asia Pacific's AI Engineering Landscape

Sea's investment in AI-native engineering is happening against a backdrop of rapid AI adoption across Asia Pacific, but with a distinctive regional character.

Southeast Asia's developer ecosystem is large and growing, but it faces structural challenges that AI coding tools are uniquely positioned to address. The region spans multiple languages, regulatory frameworks, and technical infrastructures. Developers who can leverage AI agents to navigate this complexity have a compounding advantage over those who cannot.

Other major technology companies in the region are making similar investments, though with different emphases. Grab has been an enterprise user of GitHub Copilot. ByteDance develops internal AI coding tools for its China-based engineering teams, though its Southeast Asian operations (TikTok) have a different technology stack. Gojek's engineering blog discusses developer productivity improvements but has not publicly disclosed AI coding tool strategies.

What makes Sea distinctive is the depth of its partnership with OpenAI and the explicit framing of AI-native development as a regional capability rather than an internal efficiency measure. The hackathon series, expanding across four countries, signals that Sea views AI-assisted software development not as a competitive advantage to hoard but as an ecosystem capability to cultivate.

Lessons for Organizations Considering AI-Native Engineering

Sea's deployment offers three concrete lessons for other organizations, particularly those operating at scale across multiple markets.

Lesson 1: Adoption metrics tell you about tool usage, not organizational impact. Sea's 87% weekly active usage and 73% recommendation rate are impressive, but they measure whether engineers like using Codex. The strategic value comes from the structural integration: how Codex changes the organization's ability to handle complexity that individual engineers cannot manage alone. Measure organizational outcomes (time to market for localized features, defect rates in cross-market code, onboarding speed for new engineers) rather than just tool usage.

Lesson 2: Building internal AI capabilities compounds over time. Sea's investment in Compass Max v3.5 alongside its deployment of Codex is not redundant. The in-house model provides domain-specific capabilities (Southeast Asian languages, e-commerce context) that general-purpose models lack, while Codex provides the agentic execution layer. Organizations that build both the domain-specific intelligence and the execution infrastructure create a moat that pure tool consumers cannot match.

Lesson 3: Ecosystem investment accelerates the feedback loop. By running hackathons across the region, Sea creates a larger pool of practitioners who generate use cases, discover edge cases, and develop best practices. This feedback accelerates the maturation of AI-native development practices faster than any single company's internal efforts could. For organizations with regional reach, investing in the ecosystem is a form of R&D.

What "AI-Native" Actually Means in Practice

The term "AI-native" gets used loosely, but Sea's deployment illustrates what it means concretely. An AI-native engineering organization does not merely use AI tools. It redesigns its processes, training programs, and organizational structures around the assumption that AI agents are a core part of the engineering workflow.

At Sea, this shows up in several ways. Engineers operate less as individual contributors writing code and more as managers directing AI agents on multiple tasks simultaneously. Code review incorporates AI-generated assessments alongside human review. Onboarding for new engineers includes training on how to effectively collaborate with AI agents, not just how to write code. The engineering organization's structure and workflows are being shaped by the capabilities and limitations of AI tools, rather than treating those tools as optional add-ons to existing processes.

Chen predicts "a significant transformation in engineering, with software teams becoming increasingly more leveraged as AI agents take on more operational execution work." This is the shift from IC (individual contributor) to manager that AI-native engineering implies: each engineer operates with the leverage of a small team, directing multiple AI agents across parallel tasks.

The OpenAI Partnership Dynamic

Sea's relationship with OpenAI is worth noting because it demonstrates a maturing partnership model for AI tool deployment. The relationship started in February 2025, when Shopee became a launch partner for Operator, OpenAI's AI shopping agent, in Southeast Asia and Brazil. By May 2026, it had deepened into Codex deployment, a joint hackathon series, and a public interview on OpenAI's blog.

OpenAI has even posted a dedicated AI Deployment Engineer position in Singapore, focused on helping customers adopt Codex throughout their software development lifecycle. The role involves embedding with engineering leaders, designing AI-enhanced workflows, and gathering product insights from real deployments. The Singapore location is not coincidental. It signals OpenAI's strategic investment in the Asia Pacific market, with Sea as a primary beachhead.

For organizations evaluating AI tool vendors, Sea's experience suggests that the depth of vendor partnership, including dedicated deployment support, joint ecosystem investment, and public case study development, is a meaningful differentiator. The vendor that helps you build the organizational capability, not just sell you licenses, is the one that generates real value.

FAQ

How large is Sea's engineering organization? Sea Limited does not publicly disclose exact engineering headcount, but as a company operating Shopee (the largest e-commerce platform in Southeast Asia), Garena (a leading global games developer), and Monee (a digital financial services provider), its engineering teams number in the thousands across multiple countries.

What programming languages and tech stacks does Sea use? Public GitHub repositories from Shopee's engineering team indicate significant use of Golang for backend services, along with Kubernetes operators and DevOps tooling. The front-end stack includes standard web technologies. The specific AI tooling integrations have not been publicly detailed.

Is Sea using Codex for all three of its businesses? Based on public statements, the Codex deployment appears focused on the broader engineering organization. David Chen's role as CPO of Shopee and the hackathon's association with Shopee's brand suggest e-commerce engineering is the primary focus, though the AI Centre of Excellence serves the entire company.

What is Compass Max v3.5? Compass Max v3.5 is Sea's proprietary 245-billion-parameter large language model, tailored for Southeast Asian languages and e-commerce contexts. It and its variants power AI features across Shopee's platform. The model operates at lower cost than commercial LLMs, according to Sea's public statements.

How does this compare to how Western tech companies deploy AI coding tools? The structural pattern (workflow integration, organizational restructuring, ecosystem investment) is similar to what companies like OpenAI itself reports internally. The distinctive element is the regional ecosystem approach: Sea is using its deployment to cultivate AI-native development practices across multiple countries, which Western companies operating primarily in a single market have less incentive to do.

References

"Sea Limited AI 原...
"NVIDIA Engineers...

Comment