Building Food Metadata with LLM Juries
careersatdoordash.com>Evaluators validate each tag individually — for example, protein, preparation, or health, individually rather than judging the item as a whole.
Am I reading this right that the jury is multiple LLMs each iterating through each tag and voting on each? Why wouldn't you tune one LLM to be really competent at a single tag? Like a single "spicy evaluator LLM" or "protein evaluator LLM"?
I am guessing there are so many possible tags that this wouldn't be pragmatic
I am sorry to be harsh but I find it amateurish that they would use an AI generated hero image for this and presumably fabricated LLM output -- fabricated by an AI image generator no less
Whenever I create an image like this for the purpose of a demo, I make certain that it demonstrates either real input/output or at least is exemplary of real input/output because the whole point is to instill confidence in the tool. Sure, if the raw outputs aren't clean/comprehensible enough for presenting to stakeholders or others, fine, clean them up to make them comprehensible or add explainers, but there shouldn't be any need to fabricate the inputs.
I feel obligated to respond to the hypothetical "But they don't want to tie it to a particular restaurant or brand" -- you don't have to! Doordash has taken generic food photos for this exact purpose.
Basically it’s AI on top of AI for metadata extraction.
There are a lot of claims in the article but not a lot of hard data. In the end they still don’t know if the data is correct.
Good luck with your glutes allergy.
The weird thing for me is the prompt optimization loop? Why not fine tune the model instead of AI generating the prompt?
Also, "healthy" as a boolean flag is, franky, a bit of a joke.
I see why you're confused. Some of these tags carry nutritional information, so it's easy to assume they all do.
The healthy tag is about marketing. It's a product positioning statement.
I'm guessing that if it is a real tag, then consumers have a "know it when I see it" feeling for certain kinds of food that they'd describe as "healthy." As a word that could be consistently well-defined, it's garbage. But that doesn't mean it's not useful for real-world consumers to find what they believe they want.
If I search "healthy" in DoorDash, Taco Bell comes up in the top results
> The weird thing for me is the prompt optimization loop? Why not fine tune the model instead of AI generating the prompt?
Why is it weird to optimise the prompt? Whether you optimise the model is a separate issue.
If you use any closed models you can’t fine tune them, which is another reason for most but here they also fine tuned models.
I am taking from their hero image that something like a gluten allergy would have to be verified by the merchant, but I’m just guessing that’s true. Feeding failure cases into an AI-led prompt tuning agent to solve them seems prone to a lot of problems though.
despite having multibillion dollar valuation and a real product and service doordashes tech blogs have always been surprisingly simplistic and borderline elementary or in this case, kind of just slop. I remember reading their early data science ones and they were all sort of comically limited or discussing tradeoffs between outdated methods.
Folks, not EVERYTHING needs to be LLMs!