A metabolic intelligence platform built on 10,000+ real glucose readings from continuous monitoring — not a generic assumption about what "healthy" means.
Calories matter — energy balance is what drives weight, and weight drives metabolic health. But calories only tell you the quantity. Two meals with the exact same calories can affect your blood sugar completely differently, depending on what's in them, the order you eat them, and the time of day. Wellf adds the layer calories can't see — a complement to counting, not a replacement for it.
The same meal can spike one person's blood sugar several times higher than another's. Generic advice can't see that difference — because it was never about you.
One chart for everyone. But two people can eat the identical meal and one's blood sugar rises several times more than the other's. An average can't tell you which one you are.
Essential for managing weight — but they measure quantity, not response. The same dish baked instead of fried can land far gentler on your blood sugar, a difference the calorie count alone won't show.
Good intention, no answer. It can't tell you the one thing you need in the moment: what will this plate, right now, do to my blood sugar?
Type 2 diabetes is when the body loses control of its blood sugar. Fatty liver disease is fat building up in the liver — often silently. Both are widespread across the Gulf, and both start long before any diagnosis. The figures below are shown as ranges, because credible sources disagree on the exact number.
In one national study, more than half of Saudi adults over 30 already had diabetes or its early warning stage. But the number that matters most is this: roughly 40% of those with diabetes don't yet know. This isn't a distant risk — it's here now, and largely invisible.
Wellf didn't start as a product. It started as a question its founder couldn't answer with any existing tool: after years of disciplined tracking, why was a new city, a new job, and new pressure quietly moving markers that food alone couldn't explain?
The only honest way to answer it was to stop assuming — and start reading the blood directly. So the experiment began: a continuous glucose sensor, worn through an ordinary Riyadh life, logging every meal, every meeting, every morning.
This is the pattern the whole product is built to catch. A standard panel is a single snapshot, taken once, on one morning. It comes back inside the reference range — and the story ends there. Meanwhile the continuous stream tells a different story entirely: spikes, timing effects, and stress responses that a one‑off reading was never designed to see.
Being told you’re fine while your body is already drifting is not reassurance. It’s the exact gap Wellf exists to close.
*Sampling rate depends on the sensor — this data was captured roughly once a minute.
The prediction is built in three stages, each one moving from the general population toward you, specifically. Every stage tightens the estimate.
Decades of published food‑and‑blood‑sugar research give every meal a starting estimate — from the very first photo, no sensor needed.
Real responses from users with similar bodies and diets sharpen the estimate for the food actually eaten here — not a Western database's guess.
During a sensor season, your own readings train the model on how your blood answers each food, at each time of day. That calibration then stays.
Each stage is measured, not assumed — and every prediction carries its real confidence. A new user gets an honest estimate; a calibrated user gets something close to a personal model of their own metabolism.
No food was involved in the higher spike. Stress alone did it.
Calm, unhurried, indulgent — and, on the blood, the gentler day of the two.
Forty minutes in. No food at all. The body released its own stored sugar — the same surge a burger and fries would cause.
Peak blood sugar, mmol/L · green line marks the healthy ceiling of 7.8
Two patterns a normal blood test can never catch — both obvious the moment you watch continuously.
Working out too soon after a big meal drove blood sugar too low. Waiting a few hours turned the same workout into the best result of the day. The clock mattered more than the effort — seen again and again across the data.
Every morning, blood sugar climbed on waking with no food at all, then settled within minutes. A once‑a‑year fasting test, taken at exactly that moment, could easily read this natural rise as a warning sign.
"What are you about to eat?" No food diary. No macro spreadsheet. The dial is today's impact score, the note underneath is the one thing worth doing about it.
Photo, voice, barcode, or type — whichever takes the fewest seconds in the moment. Friction is the enemy of a real dataset.
A population food‑science model predicts the likely glucose response — honestly labeled at roughly 52% baseline confidence. No sensor required.
During an optional CGM season, the model calibrates against your real readings — and keeps that personal calibration after the sensor comes off.
Every prediction tells you how sure it is. On day one that's an honest estimate. After a sensor season it becomes real precision — and you watch it get there. That climbing number is the reason to finish.
Useful from the first meal, using published food science alone — and clearly labeled as an estimate.
Sharpened by your own readings until the prediction closely tracks your real blood sugar.
Foods it doesn't recognize lower the confidence and are flagged, never quietly ignored. You always see a score with its confidence — never a falsely exact number.
A banana, logged by photo. Predicted rise: +4.7. Felt impact: light. No lab, no sensor in this moment — just the model reasoning about a real food, in real time.
Three simple, research‑backed levers — each one a small change with a measurable effect on your blood sugar.
Eating vegetables and protein before the carbs lowered the post‑meal peak by about a third in a controlled trial.
Thirty minutes of walking after a meal significantly flattened the blood‑sugar peak in trials — no gym, no equipment.
In our own sensor data, baking instead of frying the same dish produced a visibly gentler rise — same food, different result.
Sources — eating order: Shukla et al., Diabetes Care, 2015 (~29% lower peak at 30 min). Walking: Bellini et al., Nutrients, 2022 (significant peak reduction, p<0.009). Baking vs frying: Wellf's own sensor data.
Most glucose apps need you wearing a sensor forever. Wellf uses it once, to learn you — then hands the knowledge back and lets the sensor go.
You start with honest estimates from food science. Useful immediately, for anyone, with nothing on your arm.
For a defined stretch, a sensor watches how your body actually responds — and teaches the app your personal patterns.
The sensor comes off. What it taught stays — your predictions remain personal, with nothing left to wear or renew.
The proof of learning isn't a badge. It's the gap between prediction and reality closing — on the foods you eat most.
When the model is calibrated, the sensor comes off — the learning stays. That's CGM as onboarding, not obligation.
Levels and Nutrisense are the US leaders. Here's where Wellf is built differently.
| Levels / Nutrisense | wellf | |
|---|---|---|
| Wearing a sensor | Required, ongoing | Only for a short season, then optional |
| What you pay for | A never‑ending sensor subscription | Software that keeps working after the sensor |
| Where it's built for | The US market and its rules | The Gulf, and its data laws, from day one |
| The food it knows | Western food databases | Regional dishes, in Arabic, from the start |
Vision 2030's preventive‑health mandate is actively pushing early intervention over late‑stage treatment.
Saudi Arabia has some of the highest rates of diabetes and fatty liver in the world. This isn't a hypothetical market.
The US leaders aren't here. An app that understands the region's food and speaks Arabic from day one is a real advantage, not a checkbox.
Any one is copyable. Together, built Gulf‑first from day one, they're a position a US‑built incumbent can't retrofit.
Kabsa, mandi, dates, Arabic breads — understood as real dishes, not guessed from a Western database. A prediction is only as good as the food it recognizes.
Native Arabic capture, voice, and interface — not a translation layer bolted onto an English product. The market's own language, from the first screen.
Health data handled in‑Kingdom by design. The compliance moat exists before day one — not bolted on after a regulatory scare.
Saudi Arabia's Personal Data Protection Law (PDPL) requires health data to be handled in‑Kingdom under a defined legal basis. A US‑built competitor has to retrofit compliance into an architecture that was never designed for it. Wellf is built sovereignty‑first — the trust moat exists before day one, not bolted on after a regulatory scare.
The product is running on real data today. What's ahead is scale and live sync — not a first version that still has to be proven possible.
Free or low‑cost population prediction. No sensor. The acquisition layer.
A paid, time‑boxed hardware + software bundle. The calibration event — and the core revenue moment.
Ongoing personalized coaching between seasons, with optional recalibration seasons over time.
Every step removes a little more effort — until predicting your blood sugar takes nothing from you at all.
Snap a photo, speak, scan, or type. The prediction is live now — and sensor data can be added to sharpen it.
Readings flow in automatically in the background — no more uploading anything by hand.
Connected to your everyday wearables, Wellf sees the pattern coming — before you even reach for your phone.
The people Wellf is built for have already been told their labs are normal — while their body already knows otherwise. 40% of Gulf diabetics don't know they have it yet. Wellf's bet is that proof, not assumption, is what finally closes that gap.