Mixed-rice stalls feed 180 million people every day. Almost none have an itemized receipt, inventory, or a usable dataset. That's the gap.
Cashiers guess the total in 2 seconds. Same plate, three different prices.
Queues snake past lunch hour. 30% of customers walk away before reaching the till.
No itemized record. Owners can't tell what sold, what was undercharged, what was eaten by staff.
1080p ceiling-mounted unit, pre-flashed, magnetic base. 15 minutes to install — no rewiring, no scale, no kitchen downtime.
smr-vision is trained on 4.2M tagged SEA dishes. Detects every item, runs a halal & portion check, computes total in under 200 ms.
Total pushes to your existing POS — StoreHub, Loyverse, GHL — and a DuitNow / QRIS / PayNow code lights up on the customer screen.
Modular aluminum frame, ceiling or counter-mount. Ships pre-flashed with smr-vision. No IT team needed.
No app to download, no scale. Same model runs over our ceiling camera at every stall.
SmartMixedRice replaces the cashier at the front, then quietly becomes the back-office every stall has been missing.
Sits between camera and StoreHub, Loyverse, Square, BIG-POS, GHL, even legacy thermal printers.
Owners see which dishes sell, when, at what margin. Auto-flags shrinkage and time-of-day under-pricing.
Each sold portion debits the kitchen. Tomorrow's prep list lands in WhatsApp at 9pm. No more 'sambal habis' at noon.
Trained with halal-supervisor labelers. Visually flags non-halal items in halal-only stalls.
Inference runs on a Hailo-8L stick. Internet drops? Bill still prints. Cloud syncs when wifi returns.
Bahasa, Mandarin, Cantonese, Hokkien-romanized, Tamil, English. Prompts auto-localize per stall.
Mixed-rice is the default lunch across MY, SG, ID, PH and southern TH. SmartMixedRice is the first vertical AI built for its visual language.
USD · SEA economy-rice & mixed rice stall spend / yr
USD · stalls running ≥40 trays / day, Malaysia + Singapore first
USD · 2,000 stalls × RM 99–299/mo, conservative ramp
Based on pricing accuracy improvements at similar formats.
Manual cashier ~18s vs AI detection ~6s per tray.
Itemized receipt makes every charge transparent.
Product built. Seeking first pilot stalls and pre-seed.
Not a wrapper around someone else's foundation model. smr-vision is trained on SEA dish images, distilled to run on a $30 edge chip.
Hardware sold at cost during pilot phase to accelerate stall adoption and data collection.
Tiered by stall size. Starter RM 99, Business RM 299. Target 80%+ gross margin at scale.
Payment layer to be activated post-pilot. Potential to become largest revenue stream at scale.
“Reconciling every night takes hours — because there is no record of what was charged for what.”
“每天收档对账要对到凌晨,因为根本不知道哪道菜被少收了。”
“Peak hour queues snake out the door. The cashier is the bottleneck — eyeballing prices takes time.”
“高峰时间队伍排到门外,收银员靠眼睛估价,快不了。”
“Month-end numbers don't add up. No way to tell which dish, which stall, which shift is leaking.”
“月底发现钱对不上,但不知道是哪个档口、哪道菜、哪个员工的问题。”
Start free with a pilot. Scale as you grow. Hardware included during pilot period.
We are looking for our first pilot stalls. We install everything for free, you run it for 30 days. If it doesn't cut your queue and reduce disputes, we take it back. No risk.
MVP is built and working. Looking for pre-seed capital to run paid pilots, validate unit economics, and hire the first sales person in Malaysia.