Just built an AI compliance checker for K-Beauty importers. Hit me with your most brutal feedback.
Hey guys,
I’ve been dealing with cross-border trade bottlenecks and realized how much of a complete nightmare cosmetic regulations have become lately. Especially with the new U.S. MoCRA guidelines and strict EU rules, missing just one single banned ingredient on a Korean package can get an entire shipping container flagged or destroyed at customs. It's a massive financial risk for smaller importers.
To solve this specific headache for my own loop, I put together a dead-simple AI compliance engine focused strictly on K-Beauty products.
The flow is simple: you just drop an Excel sheet of the raw ingredient list or even a smartphone photo of the Korean packaging label. The system handles the OCR, translates it, and instantly cross-checks it against the latest 2026 restriction ban lists for 11 global markets in a split second.
It’s currently deployed and running live on Streamlit right now:
https://k-beauty-dashboard-cgemtdrknvzprdcxbmhk5j.streamlit.app/
I’m trying to validate if the positioning and pricing structure actually make sense to other builders. If you’re in e-commerce, logistics, or SaaS, please play around with it and give me your most honest, brutal feedback.
Does this solve a real enough pain point to justify a premium tier, or am I missing something huge? Let me know. Thanks!
This is an interesting niche problem,
from a technical perspective, I'd be curious where most errors come from today. Is it the OCR step, the translation step, or mapping extracted ingredients against different regulatory databases?
The workflow seems straightforward for users, but there are a lot of moving parts underneath, especially when dealing with multiple languages and changing regulations.
Spot on. You hit the exact bottleneck.
The OCR and translation are actually the easy parts now thanks to modern LLMs. The absolute nightmare is mapping the extracted ingredients against the regulatory databases (INCI mapping).
Korean brands love using trendy botanical extracts, but they often write them using localized marketing names. If the AI translates it literally (e.g., 'Mugwort extract'), the FDA/EU database won't recognize it because it's looking for the exact standard INCI name ('Artemisia Princeps Extract').
If it misses that mapping, a banned ingredient slips through and a container gets seized. I ended up having to build a custom dictionary fallback and prompt-engineer the hell out of the AI to force strict INCI standard outputs before it even touches the compliance database.
Have you worked with parsing unstructured data against strict DBs before? It's a huge headache lol.