ML PhD

Navigating the Shift: How to Build a Career in AI Without an Internship

Three papers in top-tier venues.

4 min readMachine Learning

A single policy change has thrown an entire career track into question. The suspension of CPT programs at major universities like UC Berkeley, Stanford, and UIUC means international students in technical PhDs can no longer take internships. For the student who posted this, with three papers in top venues like CVPR and ICRA, the worry isn't about skill. It's about access. And that distinction matters.

We understand this frustration because it points to something we talk about often: the difference between knowing how to build and being allowed to enter. This student's research in 3D reconstruction, specifically Gaussian Splatting, is precisely the kind of work that industry labs want. But without the internship pipeline, they face a double bind. Their publication record is strong, but hiring managers often treat internships as proof of cultural fit and applied ability. That's not a reflection of their potential. It's a structural barrier that has nothing to do with their capabilities.

What we would tell this student, and anyone in their position, is that the rules of the game have shifted, but the game itself is not over. Industry labs hire for demonstrated skill, and a CVPR paper is a form of demonstrated skill that no internship can replace. The key is to reframe the narrative. Instead of leading with what you couldn't do, lead with what you did do. Three first-author papers at top conferences is a signal that you can formulate problems, execute research, and deliver results. That is not a consolation prize. That is a credential.

We also want to point out that this situation is not unique to them. Many international students are navigating the same uncertainty, and the ones who succeed are often those who treat their research output as a portfolio rather than a stepping stone. If you're in this position, consider how you present your work. A strong GitHub, a clear website, and a focused explanation of your contributions can carry weight that an internship letter would have provided. And for those of you reading this who are hiring, this is a reminder to look beyond the checkbox of internship experience. Some of the strongest candidates you will ever meet are the ones who couldn't get an internship, but still managed to publish work that pushes the field forward.

There is a broader lesson here about how we evaluate talent. The system that once favored those with the right connections is being challenged by necessity. Institutions are closing doors, but the research community is not. Conferences like CVPR and NeurIPS remain open to everyone who has the skills to contribute. That is where the signal is. And that is where this student should focus their energy.

The specific thing to watch is how industry hiring managers respond to this shift. If they continue to rely on internships as a proxy for readiness, they will miss out on a generation of international talent. But if they adapt, they will find that the work speaks for itself. This student's papers are already doing the talking. The question is whether anyone is listening. We would tell them to keep publishing, keep sharing their work publicly, and trust that the quality of their research will outlast any policy decision. The internship is a means to an end. The research is the end itself.

From Machine Learning

Hey everyone, I'm an international student studying in the US. I'm on track to graduate late next year. My research is not exactly ML, it is in 3D computer vision but have decent exposure to ML as well.

In case you didn't know, the CPT program (which let's internation students do internships) has been suspended by many top universities (UC Berkeley, UIUC, Purdue, UNC, UCLA, stanford, etc). Given that there is now no way for me to do an internship, how hard will it be for me to get a job when I'm nearing graduation?

Read the original at Machine Learning