Generative artificial intelligence is already altering how students study, write and prepare for jobs. Miriam Friedel, a former vice president of machine learning engineering at Capital One who joined the University of Virginia School of Data Science faculty in 2026, lays out what students should learn about AI, what it can and cannot do, and which skills employers are seeking.
What to learn first
Friedel says the core of data science and machine learning remains the foundation: mathematical concepts and the ability to frame the right problem. “The hardest part of data science is understanding the right problem to solve,” she said. Knowing when a dataset is useful, which model choice fits a task and how to measure success are skills that AI tools do not replace.
“The hardest part of data science is understanding the right problem to solve.” — Miriam Friedel, University of Virginia School of Data Science
Key technical building blocks Friedel highlights include loss functions, linear regression and gradient descent — the fundamentals that underpin more complex systems such as transformer architectures used by large language models. She argues that students who understand these concepts will be better able to apply and scrutinize AI tools with technical rigor.
Where AI helps and where it hurts learning
- Augmentation: AI can speed up data exploration, suggest model approaches and automate routine coding tasks, freeing time for higher‑level thinking.
- Risk of overreliance: Friedel warns against taking AI output at face value. Generative models can be “sycophantic,” meaning they often echo what users want to hear rather than provide critical insight.
- Human judgment: Business context, ethical tradeoffs and problem definition remain human responsibilities. Employers still value the ability to ask the right questions and interpret results in real‑world terms.
Skills employers are looking for
According to Friedel, the most sought‑after abilities combine technical fluency and domain understanding: the capacity to translate a business need into a data question, select appropriate evaluation metrics, and validate that a model’s outputs actually serve people. She also stresses collaboration — having another person to challenge assumptions is a safeguard against poor decisions driven by polished but flawed AI answers.
Practical advice for students
- Learn the fundamentals first: statistics, linear algebra, optimization and basic modeling techniques.
- Practice problem framing: work on projects that start from a real question, not only from available data.
- Use AI as a tool, not an oracle: check outputs with independent tests and human review.
- Build communication skills: explain model limitations and impacts to nontechnical stakeholders.
Friedel’s perspective underscores a central point for students navigating an AI‑infused education and labor market: technical tools are advancing rapidly, but the human skills of judgment, context and skepticism remain essential.
