You pay per kilo we dehead. Nothing upfront. Cheaper than doing it by hand.
We at nuhuman have been building physical AI since 2025, partnering with Jetson Robotics, the IIT Kanpur incubation cell, and Coastal Corporation (NSE: COASTCORP). Our first contract: 3 cells live today (full scope: 120+ cells).
America imports 90% of its shrimp because almost none of it is deheaded there. Europe does the same: nine million kilos of Dutch shrimp go to Morocco and back every year to be peeled and deheaded by hand. The same cut costs 20 times more in Louisiana, so it sails 15,000 km instead.
Now even in India, the issue is the same: lack of labour. Top processors run lines short of workers, pay contractors on top of wages, and still miss the margin. The offshore fix is breaking at both ends.
90%
Of America's shrimp is imported
$6B
Of shrimp crosses an ocean every year
20×
The cost of the same cut in Louisiana
1.62M t
Of CO₂ a year, shipping shrimp across borders and back
Our competition is not a company. It is hand labour that is running out.
A bimanual manipulator arm for deheading shrimp. Built for one job.
Decap0 is trained and already deheading shrimp at Coastal Corp: demonstrated in an operational environment. A live export line, not a demo floor. Proven metrics today, paired with a clear roadmap to a human's pace.
Hardware cost
Arms became cheap enough to own outright. Cheap arms alone were a toy.
Model capability
Policies got good enough for wet, contact-rich work. Good policies alone had nothing to run on.
The labour
The workers who used to absorb this work ran out. Both curves landed in the same year.
Wet shrimp is not on the internet. No public dataset contains a live deheading line. Every shift on every arm trains a private adapter on our backbone, and the customer who generates it cannot hand it to anyone else.
Jetson Robotics
Our hardware partner. The arms and the edge compute they run on, built to survive a wet processing floor.
IIT Kanpur Incubation Cell
Where the policy is built. A dedicated research lab at IIT Kanpur, incubated on campus.
Coastal Corp
Where it runs. Floor access, trainers and plant data from an NSE-listed exporter, our first customer.
No capex
Nothing upfront. We own the arms; the plant pays for kilos that leave the line.
Cheaper than hand
Priced below what the same kilo costs by hand, savings from day one, verified on output.
Paid from the wage line
Our fee comes from a budget the plant already spends every month. No capital request, no committee. They approve us the way they approve labour.
Coastal's deheading spend, all three plants, per year
Before
$2.6M
1,050 people at ₹20,000 a month.
After
$1.56M
350 Decap0 arms, per kilo. 40% saved, nothing upfront.
$200
To build one arm. The plant pays the electricity
$70K
Of hardware to replace all 1,050 of them
$1.56M
A year in revenue, from one customer
A kilo of shrimp leaves Andhra Pradesh for a shelf in Texas or Rotterdam. It is deheaded, frozen, packed and shipped, and every step must be recorded: that record is what HACCP, BRCGS and import rules are. Today it is paper, easy to fake, costly to audit, slow in a recall.
The bar is rising: FSMA 204 puts shrimp on the FDA's Food Traceability List, demanding a lot code on every lot, Key Data Elements at every Critical Tracking Event, and a sortable spreadsheet in 24 hours, foreign firms included.
Under the rule, deheading is a Critical Tracking Event: transformation. The robot writes its Key Data Elements by working: quantity in and out, which is yield, grade, line and lot, stamped to the second, plus the cleaning log HACCP demands. No form filled by hand.
On-chain identity
Every cell gets a CIP-0170 identity on Cardano, each lot fingerprinted against it, on a chain nobody can edit, us included.
Private data, public proof
The data stays private; only the proof is public. QC signs every batch from their own wallet; each pack carries a QR: scan it, verify nothing changed. Paper asks for trust. This can be checked by anyone.
Identity is step one
The grant also test-runs fractional ownership of the robots on mainnet, mapping the legal path.
This is not a plan. It is already running on Cardano Preprod, and every record is public.
18 / 18
Transactions block confirmed
F1–F9
All flows represented
29 / 29
Validator scenarios passing
CIP-170
Identity authority active
01 · The stack
We own it end to end.
Arm, policy, deployment and billing are one product. Zero capex and per-kilo pricing only work if nobody in the chain needs a margin. Nobody else owns the chain.
02 · The buyer
We are paid out of the wage line.
A budget the plant already spends. No IT review, no procurement cycle. The buyer is the owner, not a committee.
03 · The data
Wet shrimp is not on the internet.
Every shift on every arm trains a private adapter on our backbone. The customer who generates it cannot hand it to anyone else.
04 · The anchor
We started at the top.
Our first cell runs on a live export line at an NSE-listed exporter, and the next buyers all know them.
→ Own the chain. Sell to the owner. Keep the data. Start at the top.