◎
See
AI Vision
Industrial cameras and tuned lighting capture each item at the most stable point on the belt.
Now piloting cashew & lime in Vietnam
AI robots that sort agriculture’s hardest products
Smart Pick replaces the hardest manual job on the line — sorting — with low-cost, high-speed AI robotic cells. Breakthrough AI vision meets breakthrough Delta robotics.
3×faster than manual
~6 motypical payback
80–110picks per minute
3 workersreplaced per cell
WW · 99%
Decay · reject
Cell 01 · Cashew lineLive
104picks/min
98.4%grade acc.
0.0%foreign matter
Live sorting timeline
08:21:03
WW kernel identified
Decay defect flagged
Robot pick & divert
08:21:05
The problem
Sorting is the most labor-intensive step on the line
Every grade holds specific commercial value. Removing defects, segregating grades, and pulling foreign matter — the step that decides profitability — is still done entirely by hand.
- Peeling is automated and color sorters only see color — fine sorting still relies on hundreds of hands.
- Skilled labor is scarce, costly, and fatigue-prone; standards drift from shift to shift.
- Over 30 million people sort by hand today — a vast, addressable wedge for automation.
A single cashew sorting floorHundreds of hands doing what AI can.
The solution
A self-contained AI cell: it sees, decides, and picks
The AI Sorter SPv.01 fuses a vision brain and a robotic arm into a single cell that drops onto the line you already run.
⚡
Decide
Grade & defect
A multi-class AI model classifies down to the defect, assigns confidence, and stores review samples.
✦
Pick
Delta robot
A high-speed Delta robot picks and places precisely — built on 7 years of parallel robotics.
Why it’s hard
One line, six classes of decision
Beyond simple color, every piece must be graded by shape, skin, oil, and defect — a multi-class problem only vision AI can solve at scale.
Whole gradesWW · LBW … intact kernels
Broken gradesSS · clean broken …
Skin-on / TestaTesta · BT · ST …
Oily gradesOW … oil-affected
DefectsSK · decay · scrap · hard shell …
Foreign matterGlass fragments · safety-critical
The product
AI Sorter SPv.01
Faster, flexible, and paid back in months. One cell replaces about three manual sorters.
80–110picks/min — 3× faster than manual
4-wayflexible AI: size, color, shape, defect
~6 mopayback, replacing 3 workers/cell
100%proprietary dataset, tuned in Vietnam
Why Smart Pick
Built different, by design
Low-cost robotics, defect-level AI, and a dataset that compounds — engineered together for SME economics.
Sees · decides · picks
AI vision and a high-speed Delta robot fused into one unit.
~$10k
Built for SMEsPer cell
A fraction of high-end optical sorters.
6
Beyond colorClasses of decision
Shape, skin, oil, defect, foreign matter.
A data flywheel
Every deployment feeds the dataset; better models ship to the whole fleet.
Designed to scale
From one cell to a fleet across Southeast Asia.
7 yrs
HeritageDelta robotics
Exported globally before Smart Pick.
3×
ThroughputFaster than manual
80–110 picks per minute.
The moat · data flywheel
Every cell makes every cell smarter
Each deployment feeds the dataset; the dataset improves every model; better models ship back to the whole fleet.
01
Capture
Production data
Real production-line data collected across factories and operating conditions.
02
Label
Web-app platform
A labeling platform that lets the community grow the dataset together.
03
Update
Edge AI
Improved models are deployed to every Smart Pick unit via Edge AI.
Competitive
Low cost meets defect-level intelligence
Color sorters can’t read defects. High-end EU/US machines price out SMEs. Manual labor is scarce. Smart Pick is the only option built for SME economics with defect-level AI.
| Approach | Cost | Defect-level AI | Flexibility | SME ROI |
|---|---|---|---|---|
| High-end optical sorters (EU/US) | High | Partial | Low | Slow |
| Color sorters (China) | Medium | None | Low | Medium |
| Manual labor | High long-term | Human | High | Scarce |
| Smart Pick AI Sorter | Low | Yes | High | ~6 months |
The mission
Thirty million people sort the world’s food by hand.
We’re building the low-cost robots that change that — starting with the crops that feed Southeast Asia.
Live pilots
Customers co-build with us in the field
Given their labor challenges, customers readily collaborate during development — they help capture and label data because the solution directly improves profitability.
Live pilot
Cashew sorting
Binh Phuoc · Smart Pick’s beachhead market.
Live pilot
Lime sorting
Long An · proof of expansion into fruit.
The market
Start with cashew, expand across the region
The same AI models and data pipeline adapt to new crops, materials, and regions — including large sorting problems like plastic waste.
Cashew→Fruit→Other crops→Plastic recycling→Across SEA
$10BTAM — cashew + fruit (SEA, India, Africa)
$2BSAM — crops & materials expansion
30M+manual sorters to replace
$10kper cell — right-sized for SMEs
Automate your hardest sorting line
Share your crop, line speed, and sorting grades. Smart Pick will recommend a pilot path to validate accuracy, throughput, and ROI on your own product.
Book a pilot