Stephanie Kirmer, a Staff Machine Learning Engineer with nearly a decade of experience in data science and machine learning, brings a unique perspective to the field shaped by her academic background in sociology and the social and cultural foundations of education. Her journey from higher education administration and teaching sociology and health sciences to becoming a key player in AI reflects a deep commitment to understanding the societal implications of technology. Her sociological training has fundamentally influenced how she approaches AI. She applies the sociological method—asking questions about inequality, differing experiences across populations, and the role of institutions—in analyzing AI systems. This lens allows her to critically examine not just how AI works, but who benefits, who is harmed, and how power dynamics shape its development and deployment. At DataGrail, where she has worked for over two years, Stephanie has seen her daily work evolve significantly with the rise of large language models. She uses code assistants powered by LLMs to brainstorm, refine ideas, and handle routine tasks like writing unit tests or boilerplate code. However, she emphasizes that these tools don’t replace the need for human expertise, especially when tackling novel or complex problems. She remains cautious about the risks and ethical concerns tied to LLMs, recognizing their potential for misuse and societal harm. When asked about the sustainability of the current AI economy, Stephanie believes a bubble is forming—similar in spirit to the dot-com era. She argues that while LLM technology has real value, the sky-high expectations and massive investments are not grounded in realistic returns. The pressure to deliver exponential profits has led to inflated promises, which she believes will eventually lead to a correction. She suggests that a more sustainable path would involve accepting moderate returns on investment rather than chasing astronomical gains. On rebuilding public trust, Stephanie points to the disconnect between AI companies’ grand promises and real-world outcomes. The cultural backlash against generative AI, she says, is partly a reaction to this hype. She advocates for more honest communication—focusing on practical, meaningful applications of AI rather than flashy, unrealistic claims. Public education about what LLMs actually are and what they can realistically do could go a long way in reducing fear and misinformation. Her writing process is organic and driven by observation. She pays close attention to how AI appears in everyday life, in news stories, and in conversations with others. She often uses sociological themes—such as power, race, class, and gender—as frameworks to explore emerging AI trends. When an idea captures her curiosity and feels urgent to investigate, she dives in. While she doesn’t plan far ahead, she’s open to new topics and invites readers to share social issues they’d like her to explore, especially those intersecting with AI. Her upcoming talk at ODSC East in April 2026 on customizing LLM evaluation promises to build on her ongoing mission: to make AI development more thoughtful, responsible, and grounded in real human needs.