
CAMBRIA January 2027: free AI safety bootcamp in Cambridge, MA
CAMBRIA is a free three week AI safety bootcamp in Cambridge, MA, from 4 to 22 January 2027, with housing, meals and travel support. Apply by 1 November 2026.
CAMBRIA is a three week program run by the Cambridge Boston Alignment Initiative (CBAI) to build technical skills for AI safety work. It is based on the ARENA curriculum. The January 2027 cohort runs from 4 January to 22 January 2027, full time and in person in Cambridge, Massachusetts.
The program is free. Participants get housing in a shared rental or hotel with their own room, lunch and dinner on weekdays, round the clock office access in Harvard Square and dedicated teaching assistants. CBAI commits to reimbursing up to $750 of travel per participant and can cover more case by case. It says it usually reimburses all travel costs.
The curriculum starts with PyTorch and ends with writing a Transformer from scratch. It then covers mechanistic interpretability, LLM evaluations, AI control and recent alignment results. The last four days are a short capstone research project with presentations.
CBAI welcomes a wide range of applicants, such as software engineers thinking about a career change, math or physics students and policy people who want to work with technical material. You need to know Python and be comfortable with multivariable calculus and linear algebra. You must be at least 18 and able to enter the US. International applicants are welcome.
Applications close on 1 November 2026, anywhere on Earth. Apply with the Airtable form linked from the program page. CBAI may ask applicants for a technical screen or an interview.
Quick answers
Can I apply from outside the US?
Yes. US work authorization is not needed. CBAI cannot sponsor a visa but can send a letter of invitation.
Can I keep a job or study during the program?
No. The program needs about eight hours of focus a day, so a full time job or school at the same time does not work.
Who is less likely to get an offer?
People with a long record of ML or AI safety research, and people who already did a similar program such as ARENA, MLAB or TARA.