I led the Discovery phase for an NHS AI / ML project that was looking to support coders in capturing what diagnoses and procedures were performed on patients in 2 NHS Hospital trusts in the midlands.

Sadly, after undertaking 1 week of primary research that 50% of the data for the ML model was dirty and thus it would never be able to equal, let alone surpass the humans. Furthermore the key insight was that there were operational and process issues driving the inaccurate data and to address those further research would be needed.