Mehrnoosh Sameki is a Principal AI PM at Microsoft, where she leads the Generative AI evaluation team focused on building and implementing Responsible AI tools within Azure AI Studio. She is a co-founder of several open-source AI ethics tools, including Fairlearn, Error Analysis, and the Responsible AI Toolbox. Beyond her work at Microsoft, Mehrnoosh serves as a curriculum developer and lecturer with Break Through Tech, helping underrepresented women pursue careers in data science.
Exaltitude newsletter is packed with advice for navigating your engineering career journey successfully. Sign up to stay tuned!
As AI reshapes industries and workflows, one role is emerging as the next big thing in tech—and it’s not software engineering. It’s product management. Specifically, AI product managers (AI PMs) are becoming the quiet force behind how intelligent systems are built, tested, and deployed responsibly. I sat down with Mehrnoosh Sameki, a principal PM at Microsoft, to learn what the job really involves, why it’s becoming so essential, and how people from all kinds of backgrounds—technical or not—can break in.
AI product managers (AI PMs) do much more than manage timelines or write specs. They work at the intersection of technology, strategy, and ethics. Mehrnoosh Sameki, a principal product manager at Microsoft, explained that her job is to make the buzzword "trustworthy AI" real and usable.
"My mission is to make sure responsible AI is operationalized for developers," she said. "Everyone cares about AI ethics, but many don’t know how to put it into practice."
As an AI PM, she doesn't just define the product vision. She and her team prototype tools, test them with real users, run A/B experiments, and even write some code to hand off to engineers. It’s a full-stack role, she says, one that requires understanding how AI works, how developers use it, and what risks it might create.
With AI advancing quickly, many experts believe product managers will become even more central. Andrew Ng has predicted that AI PMs could become more important than engineers, since AI can now do more of the actual coding.
Mehrnoosh agrees. She sees PMs becoming "full-stack builders" who work directly with AI tools, analyze data, and shape user experiences. They don’t just hand off instructions to engineers—they’re hands-on throughout the process.
As for AGI, or artificial general intelligence? While some ex-OpenAI researchers say we could reach AGI soon, Mehrnoosh says Microsoft is more focused on building productivity tools with real-world impact, not chasing science fiction.
"We use AI to take away the repetitive parts of the job," she said. "So humans can focus on what’s fulfilling and meaningful."
Mehrnoosh didn’t fall into this work by accident. She made a deliberate choice to focus on building AI responsibly.
One of her biggest challenges is keeping up with how fast AI is evolving. New risks emerge constantly, and her team must build tools and evaluations fast enough to catch those risks before they affect users.
But she doesn’t believe responsibility should slow down innovation. "Trustworthy AI isn't something you add later," she said. "It should be built into the process from the start."
Mehrnoosh's team has built red-teaming tools—a kind of ethical hacking—that attack models with biased prompts to test how they behave. They’ve also built filters to block hateful or harmful content and tools to monitor AI systems after launch.
Bias in AI comes from biased data. That’s a hard truth, but it’s not an excuse to ignore it. Mehrnoosh says developers have more tools than ever to detect and reduce bias in their models.
"It comes back to the humans building the systems," she explained. "We have to test our models, look at who they might harm, and make changes."
Her team at Microsoft has released tools like Fairlearn for machine learning fairness, and the Azure AI Evaluation SDK for generative models. These tools help teams measure bias, mitigate it, and monitor AI over time.
Many people believe you need a computer science degree or years of coding experience to break into AI. Mehrnoosh disagrees.
"Start with the tools," she said. "Get hands-on with AI solutions. Learn how they work. Then look for ways to shadow product managers or try out the role in a low-pressure environment."
Programs like MAIDAP (Microsoft AI Development Acceleration Program) help people from non-traditional backgrounds rotate through real AI projects. Meta’s RPM (Rotational Product Manager) program does something similar.
Mehrnoosh’s own path wasn’t linear. She started as a research scientist, moved into data science, and then made the leap into product management. Along the way, people told her she was wasting her degree. That she needed more experience.
They were wrong.
"PMs need a wide view," she said. "Having worn different hats is actually an advantage."
Mehrnoosh has two key pieces of advice. First, always ask for what you want. Don't assume people know how to help you. Be clear about your goals and what would unblock your path.
Second, make sure your resume clearly highlights the experience that translates to product work—even if your past roles weren’t in PM. Be extra specific about the skills you’ve built that show you can think strategically, collaborate cross-functionally, and solve real problems.
“Don’t shy away from product management just because you don’t have direct experience.” If you’re curious, driven, and willing to learn, there’s a place for you in this field!
Watch the full interview with Mehrnoosh Sameki to hear her story in her own words—and get even more insights on breaking into AI product management.
Exaltitude newsletter is packed with advice for navigating your engineering career journey successfully. Sign up to stay tuned!
