As our Content Safety engineering presence rapidly expands in Singapore, we’re looking for problem-solvers who want to stretch their skills against deep technical challenges — like building safety protections directly into AI products and generative models like Gemini. For engineering leader Matthew Bilotti, building technology that protects users requires focusing on the bigger picture — designing systems that scale globally, optimizing within cost constraints, and taking full ownership of the end-to-end solution. After years of steering counter-abuse teams on large social platforms, Matthew joined Google’s Privacy, Safety, and Security (PSS) team to help scale these critical defenses company-wide. Now, he is helping establish the team's first Asia Pacific presence in Singapore, adopting a global "follow-the-sun" model to investigate and address urgent safety issues around the clock. His advice on thriving in a fast-paced, high-stakes technical field? It comes down to a clear, shared sense of purpose. "The person who will thrive in this environment will maintain grace under pressure, keep our mission firmly in mind, and diligently execute," Matthew shares. "I always remember there are users out there depending on us to keep them safe." Ready to build what’s next for global user safety? See how you can make a meaningful impact with our growing Singapore team → https://goo.gle/4xsUtFq
An important perspective on building responsible AI at scale. Content safety requires more than strong technology it needs clear ownership, thoughtful engineering, global collaboration, and a strong focus on the people who ultimately rely on these systems. Building safety into AI from the start is critical. Google
It is encouraging to see safety treated as a core engineering discipline and not an afterthought. Building protections directly into products like Gemini is the right approach, especially as AI reaches more users across more languages and cultures. Growing a dedicated team in Singapore also makes sense, since regional context matters when protecting users at scale. The reminder that real people depend on this work says a lot about the mission.
Safety in the Generative AI era can't just be an afterthought filter slapped on top of an API—it has to be baked directly into the model architecture and inference pipelines. Balancing end-to-end safety guardrails against strict latency and compute cost constraints is easily one of the most demanding engineering problems in tech today. Establishing a 24/7 'follow-the-sun' safety engineering hub in Singapore is a massive step forward for global resilience. Inspiring direction from Matthew Bilotti and the Google PSS team! 👏🛡️
The “follow-the-sun” model is interesting because safety issues don’t follow office hours. The bigger challenge is maintaining consistent standards and handoffs across regions while still responding quickly to emerging risks.
The "follow-the-sun" model is a smart way to handle urgent safety issues, because abuse and risk don't follow office hours. As AI adoption grows across Asia, having teams closer to local languages and contexts should make protections more effective. It also shows how many career paths AI is creating beyond model building, from trust and safety to policy and operations. What skills do you think matter most for people entering AI safety roles today?
Great initiative by Google to establish a dedicated Content Safety engineering presence in the APAC region. Balancing global scale, low latency, and robust AI safety guardrails is one of the most critical engineering challenges today. Expanding these capabilities in Singapore is a strong step toward responsible AI development worldwide.
The ‘follow-the-sun’ operating model is a useful reminder that AI safety is an operational discipline, not a one-time model review. The hard part is connecting detection, incident response, evaluation and product feedback across regions—while keeping latency and cost sustainable. That end-to-end loop is what turns safety protections into a dependable service.
A powerful reminder from Matthew Bilotti that cutting-edge AI must always be anchored in user safety and social responsibility. Expanding Content Safety engineering in Singapore through a 24/7 follow-the-sun model demonstrates genuine commitment to global protection. As a daily user of Gemini, the thoughtful design and real-world reliability are second to none—it’s clear that building defenses directly into the architecture makes all the difference. Commendable leadership from Google on putting user trust at the forefront of AI innovation.
Building robust safety and alignment layers directly into generative AI systems is one of the most critical engineering frontiers today. As multimodal models become more agentic, defense-in-depth against prompt injection, adversarial bypasses, and data exfiltration cannot be an afterthought wrapper—it must be architected into the core model pipeline. At Lead Centers (https://leadcenters.in/), our engineering teams focus on building resilient LLM systems, safety evaluations, and custom AI architectures. Great to see Google scaling engineering hubs tackling AI safety head-on!
What I find particularly interesting here is the idea of building safety protections directly into AI products and generative models — rather than treating safety as something added after the fact. As AI systems become more autonomous and operate across increasingly complex workflows, safety also needs to extend beyond the model itself and into the decisions made at runtime. A production AI system should be able to answer not only “Is this output safe?” but also: What evidence supported this decision? What policy applied? What risk was identified? What authority allowed the action? What limits were enforced? And can the decision be audited and replayed afterward? That is the problem space I built ADL Trinity to address: Decision Infrastructure & Runtime Decision Governance for Enterprise AI. The next generation of AI safety will increasingly require governance at the decision layer — especially as agents move from generating responses to taking consequential actions on behalf of users and enterprises. Building safety into the AI is essential. Building governance into every consequential decision is the next layer.