Ask HN: What's the current best practice for AI/ML powered predictions?

I have a set of records which each have a set of many attributes. The records have also have a 'result' which becomes clear after an indeterminate amount of time. This result is partially but not fully influenced by by the record's attributes.

I'm looking for options to predict the record's result from it's attributes. The attributes update in 'real time', and a real-time update on predictions would be great but there is no requirement for low-latency (it's not for trading or anything like that).

I'm a software engineer, but new to AI/ML solutions. I wonder if more knowledgeable folks here could steer me in the right direction? Thank you!