TechCrunch, as part of its Women in AI series that highlights the contributions of women in AI research and industry, interviewed Sophia Velastegui. Velastegui, a member of the National AI Advisory Committee for the National Science Foundation (NSF) and former Chief AI Officer for Microsoft’s business software division, shared insights from her career in the field.
Sophia Velastegui hadn’t initially aimed for a career in AI. She studied mechanical engineering at Georgia Tech as an undergraduate. However, after joining Apple in 2009, she developed a strong interest in apps, particularly those driven by AI.
“I noticed that AI-powered products appealed to customers due to their personalized experience,” Velastegui shared with TechCrunch. “The potential for AI to improve our lives, both in small and large ways, felt limitless, and I wanted to contribute to that transformation. So, I began pursuing AI-focused projects and seized every chance to grow in this field.”
A career focused on advancing AI
Velastegui contributed to the development of the first MacBook Air and iPad, eventually being promoted to product manager for Apple’s laptops and accessories. Later, she joined Apple’s special projects group, where she played a role in building CarPlay, iCloud, Apple Maps, and Apple’s data and AI systems.
In 2015, Velastegui joined Google as head of silicon architecture and director of its Nest product line. Following a short period at audio tech company Doppler Labs, she moved to Microsoft, where she became the general manager of AI products and search.
At Microsoft, Velastegui eventually led AI initiatives across all business applications, guiding teams to integrate AI into LinkedIn, Bing, PowerPoint, Outlook, and Azure. She also drove internal projects using GPT-3, OpenAI’s text-generation model, for which Microsoft had recently obtained an exclusive license.
“My experience at Microsoft was remarkable,” Velastegui remarked. “I joined during a period of major transformation under CEO Satya Nadella. Although mentors and colleagues cautioned me against joining in 2017, seeing Microsoft as falling behind in the industry, the company was quickly making strides in AI, and I wanted to be part of it.”
In 2022, Velastegui left Microsoft to launch a consulting firm and lead product development at Aptiv, an automotive tech company. By 2023, she joined the NSF’s AI committee, working with industry, academia, and government to promote foundational AI research.
Navigating the sector
When asked about how she manages the challenges of the male-dominated tech industry, Velastegui attributed her success to the women she regards as her strongest mentors. She emphasizes the importance of women supporting one another, and perhaps even more crucially, that men advocate for their female colleagues.
“For women in tech, if you’ve been involved in transformation, adoption, or change management, you deserve a place at the table, so don’t hesitate to claim your seat,” Velastegui advised. “Be proactive in taking on additional AI responsibilities, whether they relate to your current role or a challenging project. The best managers will back you and motivate you to continue progressing. If that isn’t possible in your regular job, look for communities or university programs where you can join an AI team.”
Velastegui points out that a lack of diverse perspectives in the workplace—such as AI teams predominantly composed of men—can result in groupthink. This is why she encourages women to provide feedback whenever possible.
“I strongly urge more women to engage in AI so that our voices, experiences, and perspectives are represented at this crucial stage where foundational AI technologies are being established for the present and future,” she stated. “It’s essential for women across all industries to embrace AI. By participating in the discussion, we can influence the industry and address the existing power imbalance.”
Velastegui explains that her current work with the NSF centers on addressing critical foundational issues in AI, including what she refers to as a lack of “digital representation.” She believes that biases and prejudices are prevalent in today’s AI, partly due to the homogeneous nature of the companies creating it.
“AI is trained using data from developers, who are primarily men with particular viewpoints, representing only a tiny fraction of the 8 billion people worldwide,” she explained. “If women are excluded from the development process and aren’t offering feedback as users, then AI will completely fail to represent them.”
Balancing innovation with safety
Velastegui considers the rapid pace of the AI industry to be a “significant concern,” particularly in the absence of a shared ethical safety framework. She believes that if such a framework were widely adopted, it would enable developers to create systems quickly while still fostering innovation.
However, she’s not relying on that happening.
“We’ve never witnessed such transformative technology develop at such an unyielding pace,” Velastegui stated. “People, regulations, and legacy systems have never had to keep up with the current speed of AI. The challenge lies in remaining informed, current, and forward-thinking, while also recognizing the risks of advancing too quickly.”
How can a company or developer create AI products responsibly in today’s landscape? Velastegui advocates for a “human-centered” approach that focuses on learning from past errors and prioritizes the well-being of users at its foundation.
“Companies should establish a diverse, cross-functional AI council that addresses issues and offers recommendations relevant to the current landscape,” Velastegui stated. “They should also create mechanisms for ongoing feedback and oversight that can adjust as AI systems develop.”