Perception layer
RGB, depth, hyperspectral, X-ray, scale, line-speed and robot-pose streams fused with deterministic timing via Holoscan and DeepStream.
Platform
One autonomy layer for the whole cut floor.
Perception, planning, control, simulation and assurance — unified on a plant-edge runtime that acts on your cutting, grading, detection, portioning and packing systems.
Deployed with
Northfold PoultryMeridian PorkCascadia ProteinBaltic Fillet Co.Ardenne FoodsHarbour & Sons Seafood
Design-partner and pilot plants. Names shown are representative programme cohorts. [PLACEHOLDER]
Architecture
Five layers, one loop.
Each layer is independently hardened and independently upgradable, with signed OTA delivery to the plant edge.
Agent layer
Six specialised agents under a plant orchestrator, each with explicit tool access, approval gates and idempotent steps.
Control layer
Write-back into robotic cutting cells, portioners, rejectors and graders through OPC UA, EtherNet/IP, Modbus and vendor APIs.
Twin layer
Omniverse-based simulation of anatomy, cut sequence and yield — validating a plan before it reaches the blade.
Assurance layer
The graph linking specs, HACCP controls, frames, model versions, autonomy level, approvals and outcomes.
Fleet layer
Central model management, evaluation gating in CI, staged rollout and instant rollback across every site.
Runtime
The control loop, step by step.
-
1
Perceive
Fuse RGB, depth, hyperspectral, X-ray, scale, line-speed and robot-pose streams into one synchronised view of the carcass and the line.
-
2
Plan
Compute the cut path, blade force and seam, the grade call, the detection verdict and the portion batch — validated in the twin before the blade moves.
-
3
Act
Write back into the robotic cell, portioner, grader, rejector and MES with adaptive, per-carcass control at line speed.
-
4
Verify & learn
Measure realised yield, grade, weight and rejects, log the assurance trail, and feed supervised corrections back into training.
Capabilities
Choose where to start.
Land on the highest-ROI step for your plant, then expand across the loop.
Adaptive robotic cutting and deboning
Per-carcass cut planning that follows the seam rather than a fixed program, recovering yield that repeatable motion cannot.
- Anatomy segmentation and bone localisation
- Blade force and approach-angle control
- Bone-chip and over-trim suppression
- Twin-validated cut sequences
Vision and hyperspectral grading
Predict grade, marbling, lean/fat ratio, weight and yield consistently — no grader fatigue, no shift-to-shift drift.
- Marbling and lean/fat estimation
- Yield prediction before fabrication
- Consistent grade calls across shifts
- Grade write-back to MES and pricing
Foreign material and contamination
Fused X-ray, RGB and hyperspectral detection with deterministic reject timing and a complete evidence chain.
- Bone, metal, plastic and defect classes
- Hygiene and pathogen-risk signals
- Reject and rework routing
- Frame-level recall evidence
Weight-giveaway control
Predict weight from geometry and control the portioner so packs land on target instead of above it.
- Predictive weight from 3D scans
- Fixed-weight batching
- Per-SKU giveaway analytics
- Checkweigher closed loop
Plant edge
Latency is a food-safety requirement.
A reject decision that arrives after the product has passed is not a decision. The loop runs on-site.
GPU-accelerated cells, washdown rated
Jetson-class edge compute with TensorRT-optimised models runs inside IP69K enclosures built for condensation, temperature swing and daily caustic sanitation. Nothing in the control loop depends on a network round trip.
- Sub-50 ms reject decisions [ASPIRATIONAL]
- Sub-100 ms cut-path updates [ASPIRATIONAL]
- Deterministic fail-safe line stop
- K3s edge cluster with signed OTA
A twin that pays for itself before the cut
The protein twin simulates carcass anatomy, cut sequences and yield outcomes, then hands the optimised plan to the cell. Engineers test a new cut spec against thousands of simulated carcasses instead of a shift of scrap.
- Cut-sequence and portion-batch optimisation with cuOpt
- Synthetic rare-event generation for bone chips and contamination
- Predicted-versus-realised yield reconciliation
- CI gating: no model ships without twin validation
Performance
Engineered for line speed.
4.1%
Average saleable-yield uplift on pilot deboning lines [PLACEHOLDER]
61%
Reduction in weight giveaway per fixed-weight pack
38 ms
Median foreign-material reject decision at the plant edge
99.9%
Plant-edge runtime uptime target with fail-safe line stop
Console
Supervise autonomy, do not babysit it.
Operators see confidence, autonomy level and exceptions — and only intervene where judgement is genuinely required.
Saleable yield82.4%+4.1% vs baseline
Giveaway0.7%−61% this week
FM rejects120 escapes
Line uptime99.4%+2.2 pts
Line 4 · Front-half deboning1,412 /hr83.1% yieldAutonomous
Line 2 · Breast portioning2,980 /hr0.6% giveawayAutonomous
Line 7 · X-ray screening3,140 /hr4 rejectsAdvisory
Line 1 · Primal fabrication640 /hr78.9% yieldShadow
Illustrative console data from a pilot deboning line. [PLACEHOLDER]
Comparison
Point tools versus a system of action.
| Dimension | Incumbent point tools | Slicium |
|---|---|---|
| Scope | A single machine, camera or detector | Full cut → grade → detect → portion → pack loop |
| Adaptation | Fixed motion, fixed thresholds | Per-carcass perception and control |
| Integration | Read-only or shallow | Deep write-back into cutting, portioning and packing |
| Data | Trapped per machine | Compounding cross-plant cut-and-yield corpus |
| Validation | Trial on real product | Digital twin, before the blade moves |
| Buyer outcome | Assist a human | Perform the work, supervise exceptions |
Connectors
Pre-built for the equipment on your floor.
Cutting & deboning
- Robotic primal cutting cells
- Deboning and trimming lines
- Blade force and seam controllers
- Cell safety and E-stop interlocks
Grading & inspection
- Vision and hyperspectral graders
- Carcass grading cameras
- X-ray and metal detectors
- Checkweighers and rejectors
Portioning & packing
- Portioners and slicers
- Fixed-weight batching
- Packing and palletising robots
- Labelling and traceability
Plant systems
- MES and production monitoring
- OPC UA / Modbus / EtherNet-IP
- Historians and time-series stores
- ERP and order specs
Quality & food safety
- HACCP control points
- QMS and CAPA workflows
- Traceability and lot genealogy
- Sanitation and hygiene records
Identity & workflow
- SSO / SAML / SCIM
- Role-based approval gates
- Ticketing and on-call alerting
- Shift handover surfaces
Assurance
Every action is evidence.
Autonomy in a regulated plant is only acceptable if it is provable.
SOC 2 Type II (in progress)HACCP-alignedUSDA-FSISEU 853/2004GDPRISO 27001 (planned)
Assurance-grade audit trail
Immutable logs link sensor frames, model version, autonomy level, approvals, overrides and outcomes for every agent action — built for recall defence and model governance.
Fail-safe by design
Blade, robot and line stops are deterministic and independent of the cloud. Graceful degradation returns the cell to a safe state if perception confidence drops.
Data residency & on-prem
Run fully on-premise on the plant edge, in your VPC, or hybrid. Sensitive producers can keep every frame inside the facility.
“We evaluated four vendors. Only one could tell us what the yield would be before it cut anything.”
Start with one line. Prove the yield.
Run a paid pilot on a single processing line with a defined yield, giveaway or labour success metric. Shadow mode first, autonomy only when the numbers earn it.