January AI - Metabolic & nutrition infrastructure for developers
by•
Add meal scanning, restaurant search, device-free glucose predictions, and access to millions of verified foods to your health product. Built with the precision health demands and the simplicity modern developers expect.
Sign up, create an API key, and start building for free. Available via APIs and SDKs.
Replies
Hey Product Hunt!
We’ve spent the last few years building nutrition and metabolic intelligence inside January AI. That's AI food logging and glucose predictions that don't require a CGM. It's especially useful for people who have prediabetes, type 2 diabetes, or really anyone who cares about their metabolic health.
We developed these experiences with accuracy in mind, with patented ML models as the foundation and years fine-tuning to build something that's trustworthy and effective.
The big announcement is that we’re making that infrastructure much easier for other developers to use.
With January’s APIs and SDKs, you can add bring these capabilities into your own platform:
→ Meal scanning and food recognition
→ Natural-language, barcode, and restaurant food search
→ Food logging
→ Healthy food alternatives
→ Personalized 2-hour glucose predictions without requiring a CGM
We value the precision that health demands. Our Meal Scanner benchmarks at 86% accuracy, better than frontier LLMs, and we’ve already seen customers evaluate January against alternatives like Passio and choose us because of accuracy and coverage.
We also think nutrition data is only the starting point. The bigger opportunity is helping products understand what food actually means for an individual, and eventually bring more health context into those decisions.
The other big change with this launch is developer experience. You can now sign up, create an API key, and start building through our self-serve developer platform.
If you try it, I’d genuinely love to hear what you’re building, what’s missing, and where the experience feels harder than it should.
We’ll be here all day. Honest feedback is very welcome.
Love this. The only real challenge is that these meal scanners do not recognize keto, low-carb or meat analogs like Beyond Burgers, Impossible Burgers, etc., which leads to incorrect output and counts.
@ken_clifford That's a great point, and something we are always working to fine tune in our models. It's tricky when foods with different nutrition profiles look visually similar. Our vision model does a great job handling visual nuances. It can detect, for example if a burger is a black bean burger vs a standard hamburger, even with lots of toppings and cheese.
However, as you said, meat analogs can be nearly indistinguishable from their meat counterparts. For that reason, we make it easy for users to edit scan results with AI; the user can simply correct it using natural language (e.g. "it was a veggie burger and a keto bun") and January updates the meal and nutrition accordingly.
We think that review-and-correct step is important anytime visual recognition alone isn’t enough.
Really easy to build an app off this -- for my own tests, I've been tracking my own diet and really simple to get it all working.
@dodgeball_andrew Thanks Andrew! I'm curious to hear more about your experience. Did you build an app using January's APIs or are you tracking your diet somewhere else?