It depends on the job. Not all ML jobs involve building new low level tools. You might find it is 80% dataset classification and 20% Python code customization, with little or no statistics background required.