Developing the math took about 7 years, about 10 years ago.
The main problem is that fingerprints are soft, deformable, damaged, scarred, unclear, while cryptographic keys must be unchanging.
I have proof that it is actually impossible to transform fingerprints reliably from one finger to one single number. The main stopper here is limits between prints matching each number. There will always be more borderline cases than clear cases. The obvious thought when faced with this is to use error correction to fix that, as evident from several patents and thesises. This cannot work. The reason is the same. All these system are therefore in practice very inaccurate, while one of my inventions here is the worlds best with 1% equal error rate. The probability of accepting the wrong fingerprint when equal to the probability of not accepting the right fingerprint, is 1%. The point being that it is not actually fingerprints that are rejected or accepted, but instead cryptographic keys generated from those fingerprints. I can however do even better. I circumvent the problem, thus avoiding the impossibility of doing it.
Some time after I had developed the key math I happened to get a job in a startup working on this, probably by recommendation. I invented their system. From this I got average pay, nothing more. They brag that it is a major breakthrough in biometrics. It was also named the most innovative product of the year, on some exhibition. They do not mention the inventor, me.
The company is genkey.