Advanced Technical Fellowship
Every semester, CAIAC runs an 8-week advanced implementation fellowship on AI safety, covering topics like building neural networks from scratch¹, mechanistic interpretability², evals³. The fellowship meets weekly, with lunch and compute provided, plus about 3 hours of independent work each week.
See here for more details on the fellowship structure.
Applications for the Spring 2026 Advanced Technical Fellowship are now closed!
If this is your first semester in the club, we recommend applying to our introductory technical fellowship or policy fellowship instead.
This fellowship is primarily intended for recent graduates of the introductory technical fellowship. If this is your first semester in the club, we encourage you to apply for the introductory technical fellowship.
You don't need to have worked with PyTorch or TensorFlow before, but you should at least be fluent in Python. If you are unsure about applying for any reason, we nevertheless encourage you to apply.
Each group meets once a week for ninety minutes, with dinner provided, plus about 3 hours of independent work. We ask for your availability in the application, and will attempt to accommodate people's schedules when forming groups.
If you're already fluent in PyTorch and have research experience, you may be interested in attending our research groups or applying for club membership.
The fellowship is facilitated by CAIAC members with research experience in AI safety. This includes upperclassmen and graduate students.