Niki Parmar and the Hidden Architects: Women Transforming AI Behind the Scenes

When discussing the revolutionary breakthroughs that have shaped modern artificial intelligence, few can resist mentioning Demis Hassabis, Sam Altman, or Elon Musk. Yet for every household name who captures media attention, there exist quietly exceptional individuals whose contributions proved equally—if not more—fundamental to the field’s evolution. Among these overlooked pioneers are several remarkable women whose work has literally rewritten what’s possible in AI. Niki Parmar stands as a prime example: as one of eight core authors of the Transformer architecture, her fingerprints are embedded in nearly every large language model today, from ChatGPT to GPT-4, yet her name rarely appears in mainstream tech discussions.

The historical pattern is not new. Ada Lovelace wrote the first computer program in human history, yet how many know her name? Elaine Rich authored the first textbook on artificial intelligence, yet she remains largely forgotten in popular consciousness. The “Matilda Effect”—the systematic tendency to credit scientific achievements to male colleagues rather than the female researchers who performed the work—continues to distort our understanding of who has shaped technological progress. This is precisely why recognizing these women matters.

ImageNet and the Deep Learning Dawn: Fei-Fei Li’s Foundational Gift

The modern AI revolution has a clear genesis moment: 2012, when a deep learning network called AlexNet achieved unprecedented success in image recognition tasks. Yet few trace this breakthrough back to its true source. In 2009, Fei-Fei Li, then an assistant professor at Princeton University, proposed ImageNet—a radical idea that would reshape how researchers approached machine learning. Rather than hand-coding specific algorithms to recognize dogs versus cats, she intuited that the bottleneck wasn’t algorithmic capability; it was data.

Born in Beijing in 1976 and raised in Chengdu, Fei-Fei Li immigrated to the United States at age 12 speaking virtually no English. Within two years, she achieved fluency while demonstrating exceptional mathematical talent. She entered Princeton on scholarship and would return home almost every weekend to help manage her family’s dry cleaning business. By 2007, despite these challenges, she had become an assistant professor.

Her vision for ImageNet was both simple and audacious: create a massive, systematically labeled database of images—eventually reaching 15 million samples—that would allow researchers to compare algorithmic approaches fairly. The execution, however, required innovation. Early crowdsourcing efforts suffered from workers randomly clicking through tasks. Fei-Fei embedded control images—like pre-verified golden retriever photos—to verify worker accuracy. This quality control mechanism transformed raw crowdsourcing into reliable data production.

The impact cannot be overstated. AlexNet’s 2012 victory at the ImageNet competition didn’t happen by accident; it was made possible by the infrastructure Fei-Fei Li had built. Today, when researchers describe breakthrough moments in AI, they often ask, “Is this its ImageNet moment?”—a phrase that has become synonymous with transformative datasets. Autonomous vehicles, facial recognition, and object detection all trace their lineage to this work.

Niki Parmar: Engineering the Transformer Revolution

The wave of large language models seemed to crest with ChatGPT’s emergence, but its true origin lies in a single 2017 paper: “Attention is All You Need,” authored by eight engineers from Google. The Transformer architecture described in this paper became the foundation upon which nearly all contemporary AI systems are built. Yet remarkably, many remain unaware that one of the paper’s core authors is a woman: Niki Parmar.

Niki Parmar came from India, studying at the College of Computer Technology in Pune before pursuing a master’s degree in computer science at the University of Southern California beginning in 2013. During her undergraduate years, she discovered her passion through Andrew Ng and Peter Norvig’s groundbreaking MOOCs on machine learning and AI. “I was curious about the combined power of data, pattern matching, and optimization,” she would later recall. That intellectual curiosity would define her career trajectory.

Upon graduating in 2015, Niki Parmar joined Google’s research division, initially focused on pure research challenges. By 2017, she stood among the core architects reshaping AI’s foundational technology. On her approach to breakthrough research, she offered valuable insight: “At first, the vast amount of information around me overwhelmed me. Focusing on a specific problem and exploring it deeply with peers helps you ask the right questions.” This philosophy—depth over breadth, collaboration over isolation—would characterize her subsequent ventures.

The Transformer paper didn’t merely represent an academic achievement for Niki Parmar; it served as a springboard for entrepreneurship. She co-founded Adept AI with fellow author Ashish Vaswani, who was listed as the paper’s first author. The company raised $350 million—a substantial vote of confidence in their vision. Recognizing her own focus areas, Niki Parmar also co-founded Essential AI, managing it as her primary venture. Essential AI raised $56.5 million in funding from technology heavyweights AMD, Google, and Nvidia, validating the founders’ strategic direction within the AI infrastructure and research community.

Daniela Amodei: Anthropic’s Safety-First Co-Founder

While media coverage of Anthropic frequently emphasizes that it was founded by “seven researchers who left OpenAI,” this framing obscures a crucial fact: Daniela Amodei, the company’s president and co-founder, has been systematically downplayed in coverage despite being instrumental to its vision and strategy. Anthropic was actually founded by two siblings—Daniela and Dario Amodei—whose complementary skills created an unusually balanced leadership structure.

Daniela’s professional journey reveals an atypical path for a technology executive. She earned bachelor’s degrees in English literature, politics, and music literature—a humanities-focused education that informed her subsequent focus on human values alignment in AI. Her early career in politics and nonprofit work developed her strategic thinking and organizational capabilities. In 2013, when Stripe was still a relatively unknown startup founded just three years earlier, Daniela joined as an early employee, eventually building it into a powerhouse (now valued above $50 billion at its peak).

At Stripe, she took on roles that would later prove essential at OpenAI and Anthropic: team recruitment, risk management, and cross-functional coordination. She led teams analyzing over 7,000 potential fraud and policy violation cases annually, achieving a 72% reduction in loss rates—bringing the company to its historical lowest. This operational excellence and risk discipline would become her hallmark.

In 2018, she joined OpenAI as Vice President of Safety and Policy, working not only on technical safety teams but also overseeing human resources, recruitment, learning and development, and DEI initiatives—serving as a true generalist in a specialized field. In 2021, she co-founded Anthropic with her brother, bringing this safety-first philosophy to a new organization explicitly built around the principle that AI systems must align with human values.

Mira Murati: OpenAI’s Quiet Technology Leader

OpenAI’s Chief Technology Officer since 2022 is Mira Murati—a distinction that many in the tech world remain unaware of. Mira joined OpenAI in 2018, was promoted to Senior Vice President overseeing research, product, and partnerships in 2020, and has since shepherded the development of ChatGPT, DALL-E, and GPT-4.

Born in Albania in 1988 and educated in Canada, Mira’s background is rooted in engineering. At Dartmouth College, she distinguished herself by building a hybrid race car as part of a school project—an early indicator of her hands-on, problem-solving mindset. After a brief tenure in aerospace, she joined Tesla as a Senior Product Manager for Model X, where her exposure to Autopilot kindled a deep interest in artificial intelligence.

On the subject of intellectual motivation, she once observed: “Boredom is a powerful motivator for pursuing and exploring the frontiers of anything.” This philosophy has guided her trajectory at OpenAI, where she remains deeply involved in the company’s most ambitious projects. The development of ChatGPT, perhaps OpenAI’s defining achievement, proceeded under her technical leadership. In 2023, when Microsoft committed $13 billion to OpenAI—a partnership Murati negotiated and managed—CEO Satya Nadella publicly praised her for demonstrating “the ability to assemble a team with technical expertise, business acumen, and a deep understanding of the importance of the AI mission.”

Her influence extends beyond product development. During internal crises—including leadership conflicts that threatened the organization’s stability—her voice on critical matters carried significant weight. Yet unlike some peers who found themselves marginalized, Mira has maintained her standing within OpenAI, continuing to shape the company’s technical direction and strategic decisions.

Timnit Gebru: The Ethicist Who Refused to Remain Silent

The recent decision by Google to withdraw its Gemini text-to-image model due to AI ethics concerns echoes an earlier, more dramatic clash: the 2020 dispute involving Timnit Gebru, then an AI researcher at Google, who publicly criticized the company for what she described as her termination in retaliation for raising concerns about algorithmic bias.

Born in 1983 in Eritrea and Ethiopia, Timnit Gebru completed her Ph.D. in electrical engineering at Stanford in 2014, specializing in computer vision and machine learning. Rather than pursue the conventional path of optimizing model performance, she dedicated herself to researching fairness, accountability, transparency, and ethics in AI systems.

Her breakthrough work demonstrated that commercial facial recognition systems demonstrated significantly lower accuracy when identifying women and people of color—a discovery with profound implications. Her research directly influenced Amazon’s decision to discontinue its facial recognition service, Rekognition, illustrating how rigorous ethics research can drive real-world corporate accountability.

In 2020, Gebru co-authored a research paper criticizing large language models for their environmental impact and lack of diversity in their development processes. Google’s AI leadership rejected the paper’s publication, asserting it “did not meet our publication standards.” During a subsequent conflict with the company, Gebru’s corporate email was disabled while she was on vacation—a move that ignited international backlash. Over 1,500 Google employees signed a petition supporting her, joined by more than 2,000 external researchers, nonprofit leaders, and industry peers.

Despite this unprecedented show of solidarity, Gebru ultimately departed from Google. Rather than fade from the spotlight, she established DAIR (Distributed AI Research Institute), an independent organization explicitly designed to counter the outsized influence of large technology companies in AI research and deployment. On her mission, she stated plainly: “I cannot wait for big tech companies to finally solve the problems brought by AI.”

The Broader Picture: Why Recognition Matters

The achievements of these five women—Fei-Fei Li, Niki Parmar, Daniela Amodei, Mira Murati, and Timnit Gebru—represent far more than individual success stories. They illustrate a systemic pattern: for every acclaimed male technologist receiving outsized media attention, numerous women of equal or superior capability work in comparative obscurity, their contributions systematically undervalued and their voices frequently marginalized.

Interestingly, there is a lineage connecting these figures. Timnit Gebru has worked under the mentorship of Fei-Fei Li—a reminder that progress often builds itself through generations, with earlier pioneers creating pathways for those who follow. Yet this cycle remains fragile and insufficient. Structural barriers—unequal investment access, inadequate mathematical education for girls, workplace discrimination—continue to suppress female talent at scale.

No single article can dismantle these systemic inequities. This is precisely why International Women’s Day persists, and why initiatives supporting women in technology remain essential. But on this day, take a moment to remember these architects of AI. Give credit where it is genuinely due.