Much attention has been paid to speeding up code through vectorization. The ability to gain a constant time speed-up by writing code in a way that allow it to be run in parallel or compiled down to a higher-performance language has broad applications from scientific computing to machine learning. One of the most popular libraries for writing code in this paradigm is the NumPy Python library, which defines data types and functions that allow for writing code that operates over arrays (instead of individual values) that is automatically vectorized, resulting in large speedups over code written in standard Python. However, it is often difficult to write such code, requiring knowledge about many different NumPy functions; moreover, NumPy code can be significantly less transparent than the corresponding standard Python code. In this extended abstract, we present ongoing work on a system to automatically translate code from standard Python code operating over lists into vectorized NumPy code. To accomplish this, we leverage the technique of verified lifting by building on an existing tool, Metalift, and address challenges that arise in applying existing verified lifting approaches to the domain of vectorized NumPy code.