Two high schoolers built an app that counts calories from a photo. It hit $40 million a year — and MyFitnessPal bought it.

calai.app

$40 million in trailing revenue, an acquisition by the category leader, and the founders were still writing code between high school classes when it started. Zach Yadegari and Henry Langmack built Cal AI — an app that turns a photo of your plate into a calorie count — in May 2024. Eighteen months later, MyFitnessPal bought the company outright.

Yadegari and Langmack built Cal AI from Yadegari's parents' house in Roslyn, New York, while still in high school. Photo-based calorie logging wasn't a new idea — MyFitnessPal itself has offered a version of it for years. What Cal AI got right was the accuracy and the friction: point the camera at a plate, and a multimodal AI stack — combining vision models with a phone's depth sensor to estimate portion size — returns a macro breakdown against a database of over a million foods, in seconds. No search bar, no manual entry, no guessing.

The product was the easy half. The harder, less-discussed half was distribution, and this is where Cal AI diverges from the typical bootstrapped-solo playbook. The founders started with a $2,000 test budget on TikTok and Instagram, then scaled it into a disciplined paid-creator operation — roughly 250 creators on monthly retainer, continuously producing short-form videos of the same core moment: point, snap, see your calories. That paid engine, not organic virality, is what pushed the app past a million downloads and kept it climbing.

The AI wasn't the hard part — recognition models like this already existed. The harder decision was choosing to spend real marketing money as teenagers, before anyone told them that's what serious companies do.

Revenue climbed fast and stayed verifiable at every stage: roughly $12M annualized by late 2024, over $30M in 2025, and more than $40M in trailing 12-month revenue by the time MyFitnessPal's acquisition closed in December 2025 (announced March 2026). The team grew to around 17 people, still without raising outside funding. Fortune reviewed the company's financial records directly; TechCrunch and CNBC independently confirmed the trajectory — this isn't a founder's self-reported number, it's press-verified.

1. Wrapped an existing AI capability, not a new one. Food-recognition vision models already existed. Cal AI's edge was combining that recognition with a depth-sensor portion estimate and a fast, frictionless onboarding — not inventing new AI.
2. Treated marketing as an engineering problem. A $2,000 test budget became a retained network of roughly 250 creators, run like a media-buying operation rather than a hope-it-goes-viral bet.
3. Picked a category with a built-in, tedious competitor. Manual food-diary logging (the MyFitnessPal default for over a decade) was the exact friction Cal AI removed — the comparison sold itself on first use.
4. Let the numbers do the talking to a buyer. A clean, independently-verifiable revenue trajectory — not hype — is what made a public-facing category leader acquire the company outright within 18 months.

1. A camera plus an existing AI model can replace an entire category of tedious manual entry. Food logging, receipt scanning, plant ID — anywhere a person currently types data by hand, a vision model can often just look and tell them.
2. Paid distribution, run like a real media-buying operation, can beat organic virality. Cal AI's creator retainer network was a deliberate, budgeted system — not a lucky viral moment.
3. Age and experience are not the gatekeepers people assume. Two 17-year-olds with no funding and no industry track record built something a public-facing category leader chose to acquire outright.

Revenue figures tagged self-reported come directly from the founder's public posts or interviews — we don't audit them. Where a figure is independently reviewed by outside press, as with this issue, we note that instead. Our purpose is to share success stories with enough online proof to be worth your attention, not to certify the numbers.