Nobody in this industry evaluates potential by course syllabus or completed assignments. What opens doors is evidence that you understand how machine learning leaves the laboratory and creates value in the real world.

By the end of your second year in the BS in AI and ML Engineering, the goal isn’t just to write code, it’s to demonstrate that you can think across three distinct dimensions of modern technical problem-solving:

1. Solving unglamorous, high-friction problems Most real economic value lives in the overlooked corners of industry: paperwork bottlenecks, manual data reconciliation, compliance checks, and operational friction. When you build an autonomous system that takes messy real-world documents, extracts truth, catches edge cases, and leaves room for human judgment, you prove business empathy. You show employers that you don’t just train models, you understand workflows, reliability, and enterprise value.

2. Bridging messy perception to instant utility Consumer technology thrives on reducing friction to zero. Turning an unpredictable physical world, captured through a phone camera or sensor into clean, actionable insight is deceptively hard. Building an intuitive tool that does one small visual task exceptionally well demonstrates product sensibility. It proves you understand low-latency systems, edge UX, and the discipline of creating something someone would actually rely on in their daily routine.

3. Grappling with scientific ground truth Building consumer tools and workflow automation proves velocity, but tackling deep technical domains proves rigor. Working with complex physical, biological, or chemical datasets forces you to confront noise, distribution shifts, and genuine uncertainty. When you model natural phenomena and clearly communicate where your predictions hold and where they fail, you demonstrate scientific maturity. That kind of intellectual discipline is rare—and increasingly critical as AI moves into the physical sciences.

The point of a portfolio year

A portfolio is not a vanity exercise. It is the interview. By the end of year two in the BS in AI and ML Engineering, you want three things live on a public URL, with a GitHub repo and a one-page write-up. That is the stack that gets you into the room with Harvey, Isomorphic, or the next Cal AI.

Open the brochure, look at the second-year build modules, and start sketching which three products you want on your name.