I don't think it's fair to assume that genders and ethnicities should match either the general population or the hiring pool. Geography plays a huge role: if you have a company in a city with a substantially higher asian population (like Seattle or the Bay Area) that geographic distribution is going to have a significant influence. Likewise, a product used primarily by women might have higher representation of women because of the userbase. Pinterest is a good example here.

I think the solution is to realize that it's wrong to pick outcomes a priori, that we shouldn't trying to find the "correct proportion". Instead, focus on eliminating discrimination and biases in the hiring pipeline. Anonymize the gender and ethnicity of applicants. Use voice altering software to prevent the ability to discern between genders when interviewing. If the goal is to eliminate discrimination on the basis of protected characteristics, create a hiring process that isn't able to distinguish between these characteristics in the first place. Then, whatever proportion that results is the "correct proportion".