I agree, but the definition alone isn't sufficient to actually calculate eigenvalues. Hence the standard approach which says that for matrix A, vector v, and eigenvalue λ, we have
Av = λv
=> Av - λv = 0
=> (A - λI)v = 0
=> det(A - λI) = 0
Which then yields the characteristic polynomial. Skipping the determinant means you need a different approach.
You can prove many fundamental results of Linear Algebra with the definition that does not directly use determinants. In fact, one would define trace and determinant as sum and product of eigenvalues. Definition of characteristic polynomial would then follow.
If "computation" is what you are after then Av = λv is about solving a system of equations and you can try elimination, etc.
