Velorific: Built an AI-assisted nutrition layer for Strava data after realising training load and food targets should never be separate

Hi all,

Like many of you I have been thinking about how to make training data actually work harder for me, and I wanted to share something I have been building that takes a slightly different angle to what is usually discussed here.

A bit of background. I come from an endurance running and cycling background and I work in data analytics. For years I trained seriously, tracked everything on Strava, followed structured plans, and completely winged the nutrition side. Ate by feel, assumed it was roughly right, wondered why recovery was always a bit slower than it should have been and why I could never quite string good weeks together consistently.

Turns out I was under-fuelling on hard days without realising it. The signs were all there but I blamed the training load rather than the plate. When I finally started paying proper attention the difference was bigger than I expected. FTP moved from low 260s to high 290s over a couple of months, lost 1.4kg while eating more than I ever had, recovery noticeably better.

The problem was nothing out there did what I actually needed. Every nutrition app gives you the same static macro targets regardless of what you did that day. A rest day and a four hour hard ride look identical to MyFitnessPal. That makes no sense for anyone training seriously.

So I built Velorific.

It connects to Strava, reads your training load after every session, and sets your macro targets for that specific day. The carbohydrate targets follow the Burke et al ranges from the Journal of Sports Science, bucketed into easy, moderate, hard and very hard using relative effort thresholds calibrated against real ride data. Protein sits at 1.6 to 1.8g per kg in line with the ACSM position stand. The day after a very hard session the deficit gets waived entirely based on the IOC RED-S consensus guidelines.

The design principle is deterministic, auditable, and grounded in source data rather than guesswork. The targets are computed from your actual Strava data every single day, not inferred from a static profile you filled in once at signup.

It started as a spreadsheet, then a local server on my laptop, then something polished enough to share properly.

Still early days and genuinely open to feedback from people who train seriously and think carefully about this stuff. The intervals.icu community is exactly the kind of audience I had in mind when I built it.

iOS only for now. Search Velorific on the App Store or happy to answer any questions here.

Adam + Sophia