[humblefactory]

I agree with your assessment that this idea is naïve. The sort of biases that you referred to are subtly and pervasively distributed through all of the different schematic and ideological methods that we use for choosing language. This means that your filter list would be untennably large and difficult to construct because of how broad it would be. Moreover, these sorts of biases often are based on less than liminal evaluation metrics. It's not just that hiring companies look at POC candidates and say "nope" based on a name or photo. Instead, they are drawing from pools of candidates like high-end computer science programs, which often have already done the work of exclusion all ready, because of costs or admission requirements. Or they are basing their hiring decisions on having prior experience in internships, which often don't pay enough for the high cost cities in which they are located to support folks without additional means (from parents or relatives) to manage living expenses.

Long story short, if you really want to address this problem (or implicit bias in hiring) there is actually a huge literature coming out of business schools, public administration schools, and our of corporate DEI initiatives. Software driven support for these efforts is a crucial part of moving them forward, and you should definitely spend some more of your weekends brainstorming on this!