The need to evaluate cement behind multiple casing strings has become critical in the context of well plug and abandonment (P&A) or slot recovery. A particularly important case, addressed in this work, is the application to optimize P&A operations in offshore wells where the capability to evaluate barriers behind casing using tools deployed inside an inner tubing would reduce the costs associated with rig time and potentially (in case the presence of barriers outside the casing is confirmed) the unnecessary removal of casing strings. We present an approach using a combination of sonic and ultrasonic tools deployed inside tubing to characterize the near well bore environment and in particular, the annular fill and bond behind the outer casing, i.e. in annulus B. Extensive modeling and experiments were used to characterize the sonic response in a dual string geometry and identify features that inform on the state of annulus B. Some of the complexities and ambiguities in this complex scenario are addressed by using the ultrasonic tool to characterize the annulus behind the tubing (annulus A) as well as the tubing location inside the casing. Additionally, we combine the single feature answers with those using deep machine learning to learn the complex relationship of the subtle features in the through tubing sonic response to the finer variations of the annulus B state as indicated by ultrasonic cement maps acquired in the outer casing. We test and validate these answers on a number of field jobs and show the results of the through-tubing logging answers with the reference cement evaluation maps obtained after removing the tubing.