I agree with the author that longevity biotech is frequently tripping over itself with enthusiasm to bring interventions to market (or some variety of off-label use) well before they're on solid scientific ground. You don't see this in, e.g., cancer research- if a novel kind of anti-cancer drug tests well in mice, you (almost) never see a sudden cottage industry pop up around it the next week. Obviously part of the difference is that most of us aren't expecting to suffer from cancer in the immediate future, whereas everyone expects to age; aging research faces a lot of weird psychological pressures that most industries do not.
For all that, though, there are a lot of parts where I think this article is too skeptical. As someone who uses machine learning in (non-aging-related) biotech, the discussion of "garbage-in/garbage-out" put me in mind of "Yes We Have Noticed the Skulls" [0]; we're well aware that data quality coming out of molecular biology is a problem, and anyone working in the field could give you a long list of the kinds of biases, distortions, and gaps that are known to occur in their areas of interest [1]. Not all of the grand claims made about biotech AI/ML are taking this expertise into account, certainly, but certainly many of them are.
[0] https://slatestarcodex.com/2017/04/07/yes-we-have-noticed-th...
[1] The list, regardless of what variety of data it applies to, would always end with "sometimes something unknown happened and the data just turned out bad, oh well."