SHEM
An app that reads your grocery receipts and runs your kitchen inventory.
For people who are tired of buying tomatoes they already have.
Three things, in order.
It reads receipts.
Paper, screenshots of online orders, or email confirmations. The OCR is tuned for grocery line items, so it handles the ones general OCR breaks on. Faded thermal print, abbreviations like "OG BABYSPINACH 250G", prices that wrap onto a second line.
It keeps a real inventory of your kitchen.
Items are sorted into fridge, freezer, pantry, and counter, with rough expiry windows. So the milk you bought on Tuesday flags itself by next Wednesday.
It predicts what you're about to run out of.
Based on how fast you've finished things in the past, not a generic average. If your home finishes a carton of eggs in six days, that's what SHEM uses.
People who do most of their grocery shopping online.
And lose track of what's already in the freezer. Households where two people shop and nobody knows what's in the pantry. Anyone whose fridge has had a half-empty jar of olives in it for six months.
All three inputs. No manual typing.
Most pantry apps make you type every item in by hand. SHEM is the only one we know of that handles all three inputs, paper receipts, screenshots of online orders, and forwarded email confirmations, with no manual entry for a typical weekly shop.
SHEM is in private beta.
We'd rather show it on a fifteen-minute call than post numbers that aren't real yet. Email info@ithynx.com and we'll add you to the next wave. Or walk you through it live.
Get on the beta list →For the engineers reading.
The receipt reader is a vision model trained on a few thousand annotated receipts from Indian and international grocery retailers. The prediction layer is a per-household consumption model that combines purchase frequency with category-level priors.
Photos are processed in our own cloud, parsed within seconds, and deleted within 24 hours. The extracted inventory data stays on your account. We don't sell it, we don't share it, and we don't train general models on it.
We build this kind of thing for other teams too.
SHEM is the version we built for ourselves. If you want something in the same shape, a receipt reader, a document parser, a vision-plus-prediction layer inside your own product, we take on a small number of those each year.
Email info@ithynx.com with the actual problem.