LiftEngine started from a specific frustration with existing workout apps: almost all gate the actual plan behind a sign-up or subscription, and if you do get through then the "personalised" plan is invariably a fixed template. At best they may have a few exercises lopped off to meet time/equipment constraints. I wanted something that took my real constraints - equipment, training days, session length - and produces a plan that fits whilst following a science-based methodology. The approach I settled on comes in two parts. A constraint solver handles the hard rules (recovery windows, equipment you own, time budgets) so the output is always feasible. A genetic algorithm then optimises the feasible plans against training-quality criteria I pulled from the literature. The GA part was probably the most challenging. This initially utilised the (excellent) eaopt library, but ended up adopting a custom GA engine tailored to the problem that brought memory usage per request down to ~25 MB and cut plan generation to ~1200ms. This is a side project and I'm still iterating. Happy to hear what works, or doesn't, as well as any feature requests.

Forum Threads

p/liftengine

•

4mo ago

Exercise Dataset Attribution

Product Hunt shoutouts focus primarily on products so I'd just like to use this thread to credit the authors of and contributors to the public domain dataset this project is based on.

The original dataset is courtesy of OllieJennings which can be found here on Github, it was restructured by yuhonas in this fork which I made use of.

View all

LiftEngine

Workout plans tuned to your goals, kit and schedule

Workout plans tuned to your goals, kit and schedule

Forum Threads

Exercise Dataset Attribution

Forum Threads

Exercise Dataset Attribution