Prompt to Prod: Engineering an Autonomous SDLC at Scale

Interview:

What is the focus of your work these days?

I lead Engineering Acceleration and Core Platforms at Roblox, where my team is responsible for the foundational API and compute stack serving over 144 million daily active users. My current mandate is the AI-Native Transformation, moving beyond simple coding assistants to build robust agentic infrastructure. We're focused on utilizing our deep institutional memory to automate the hard problems of scale: complex codebase migrations, autonomous maintenance, and the creation of a self-healing software development lifecycle.

What is your session about, and why is it important for senior software developers?

The session focuses on the evolution of Engineering Acceleration through the lens of agentic infrastructure. We explore how to move beyond simple chat interfaces to give AI tools programmatic access to internal APIs and development stacks. For senior developers, this is a fundamental shift in how work is assigned and executed. It is no longer just about generating a snippet of code. It is about architecting systems where AI can autonomously handle complex migrations and maintenance tasks while adhering to strict security controls and infrastructure policies. The session provides a framework for senior architects to lead this transition without sacrificing the integrity of their systems. We will also discuss how we moved the conversation toward scaling expert judgment through agentic guardrails, allowing senior architects to remain the stewards of quality while leveraging automation to eliminate the toil of codebase maintenance.

What is the motivation behind your talk?

After leading engineering teams at Google, Instagram, and now Roblox, I've seen the same pattern repeat: speed without safety just creates debt faster. The motivation for this talk is to address the Productivity Paradox head-on, the reality that while AI can write code faster, the bottleneck has shifted to review, safety, and architectural alignment. I want to provide a technical blueprint for how large enterprises can bridge this Trust Gap by grounding AI in their own historical telemetry and expert judgment.

Why is it critical for software leaders to focus on this topic right now?

Leaders are currently facing a high stakes trade off. They want to unlock the full functionality of agentic tools, but they must do so without introducing existential security risks to the company. Issues like prompt injection or a rogue agent accidentally compromising system integrity through data deletion are real concerns that can paralyze adoption. Right now, leaders must define the governance and infrastructure layers that allow AI to be truly productive. This means building a secure bridge between the AI capabilities and the company m infrastructure, ensuring that policy updates and security guardrails are applied at the speed of the agent.

What's one thing you hope attendees will implement immediately after your talk?

The transition toward autonomous engineering requires addressing fundamental infrastructure and development workflow issues. At Roblox, we have bridged this gap by evolving beyond simple generative assistants into domain aware agentic engineering. This is achieved through the implementation of AI-first novel knowledge management systems, programmatic access to internal infrastructure APIs governed by real time policy updates, and a system of Exemplar Alignment that digitizes decades of expert intuition and institutional memory.

What are the common challenges developers and architects face in this area?

The most significant hurdle is a convergence of the context gap and infrastructure governance. Most AI models have not lived through years of a specific codebase or seen the internal review standards that define an organization. Beyond that, giving AI agents secure, programmatic API access to the internal stack is technically and legally complex. This creates a security productivity paradox. AI might help a developer move faster, but without deep understanding of security layers, it could introduce a latent regression or a critical vulnerability. Digitizing the intuition of senior staff into a format that can govern these new agentic work assignments is a challenge that many teams are just beginning to navigate.

Who is your session for?

• Engineering Executives (CTOs, VPs, Directors): Architecting the long-term AI strategy for mature, large-scale organizations.
• Staff and Principal Engineers: Responsible for maintaining architectural integrity, security, and performance in complex, heterogeneous systems.
• Platform and DevEx Leads: Looking for empirical frameworks to measure and accelerate the end-to-end software development lifecycle.
   

What makes QCon stand out as a conference for senior software professionals?

QCon stands out because it prioritizes the how over the hype, focusing on the pragmatic realities of building systems that serve billions of users. Throughout my career leading teams at Google, Instagram, and now Roblox, I have found that the most valuable technical growth happens in practitioner led environments where we can dissect the difficult intersection of infrastructure, security, and developer velocity. My session on Engineering Acceleration is a direct result of these honest deep dives into the complexities of agentic infrastructure and the evolution of technical management.