I totally see your point (many competitive runner never posts to Strava, anyway). I'm working on a Garmin integration now, so that should be available soon. It looks like runkeeper only has manual activity exports, so I'd have to provide a way to allow for manual activity uploads, too, which is doable!
My Garmin times are also a joke, which is why I created getfast! (To be clear though, Garmin makes incredible watches, and their predictions have gotten a lot better in the past year . . .)
I can't say to much because we are trying to keep our methods stealth, but I think it's better to be last than first in this race. Strava isn't the only company in this field: Garmin, Kaizen, and AI Endurance are just a few. For a while, our race predictions were a lot more accurate than the biggest of these (Garmin, which has a lot more data than Strava), which is probably telling of the difficulty. There isn't an obvious company you go to to tell you about your fitness, but if there was we would have never started our mission.
I wish you luck, because I think this space is about to get a lot more crowded.
That said, I suspect that Strava and Apple (and likely even Google) have more data than Garmin. Maybe quality of data is lower (garmin is higher % athletes?).
There are a lot of companies planning or starting to tell you about fitness. Oura and Whoop and now Fitbit are happy to give you basic training readiness info. Google will help you plan runs, Apple will give you “training load”.
I’m not sure what your target market is, but don’t forget to look broadly. Garmin is favored for training athletes but fitbits and Apple Watches are favored for casual workouts. If you’re trying to train models on metabolic or other physiological training ability, don’t forget to look at non-athletes or early-athletes.
Thanks! In my experience in these things, there are only handful of people with the right ML background implement the products well (it tends to be a blend of domain knowledge, researching ideas in the correct way, and engineering to make the iteration loop fast), and I think our team has that. Hitting the right audience like you mentioned is going to key, too.
