Deploy robots where touch matters.

Monty brings physical AI into production for tasks that depend on contact, force, dexterity, and real-world interaction. We combine the sensing, data, model adaptation, evaluation, and fleet infrastructure required to make those tasks deployable.

Vision tells a robot where to act.
Touch tells it how.

Many of the tasks still beyond automation fail at the moment of contact. Handling deformable materials, regulating grip force, inserting components, and manipulating delicate objects require physical feedback that vision alone cannot provide.

A robot gripper with soft tactile fingertip pads lowering a machined steel pin into a close-tolerance bore in an aluminium fixture
VisionObjectPositionGeometryMotion
Vision + touchForceSlipPressureDeformationGrasp stability
Applications

Contact-rich work across the physical economy.

Two robot arms with tactile fingertips drawing navy fabric flat across a sewing table in a garment factory
Apparel
Deformable material handling
A multi-fingered robot hand with soft tactile pads seating a flexible ribbon cable into a connector on a circuit board
Electronics
Precision assembly and manipulation
A soft robotic gripper closing gently around a ripe strawberry still on the plant
Food & Agriculture
Delicate and variable objects
Two robot arms lifting a soft poly mailer out of a tote of mixed parcels on a warehouse conveyor
Logistics
Unstructured item handling
A robot gripper lowering a machined steel pin into a close-tolerance bore in an aluminium fixture
Manufacturing
Contact-sensitive production
Two robotic hands with soft tactile fingertips easing a knitted cardigan sleeve over the arm of an older person seated in a wheelchair
Care
Assistive physical tasks

Start with the tasks you want to automate.

Show us the workflow. Monty evaluates the physical requirements, operational constraints, and deployment environment to determine the right path to automation.

We work backward from the workflow to determine the right robot, sensing, data, and deployment approach, validate it under real production conditions, and build toward scaled deployment.

AssessCaptureAdaptEvaluateDeployOperate
Stage 01

Understand the workflow.

We evaluate the task, physical requirements, failure modes, environment, throughput, and operational constraints, then determine the appropriate deployment approach.

Stage 02

Capture the right data.

We collect synchronized demonstrations and production data, then structure, visualize, annotate, and curate the highest-quality examples for training.

Stage 03

Adapt models to the task.

We post-train models against the workflow, data, and robot embodiment to turn general capabilities into task-specific behavior.

Stage 04

Test against the real workflow.

We evaluate performance against production criteria, edge cases, failure modes, and the conditions the robot will encounter on-site.

Stage 05

Move into production.

We install the system on-site and establish the appropriate level of supervision, human fallback, and operational integration for the application.

Stage 06

Operate and improve.

Monty monitors system health, task performance, failures, and interventions across deployed robots as production conditions change.

A manual garment line: cut fabric panels, calipers and sewing stations before any automation
A worker in sensored gloves folding fabric under a camera rig, recording demonstrations
A two-armed robot folding fabric panels at the cell, feeding a conveyor
A row of robot cells running the folding task across the production floor
Monty data layerSynchronized · ±80 µs
Tactile
Robot state
Action
Adaptation · Garment handling / Line 04Running
post-train · v2.4.1step 12 400 / 20 000
post-train · v2.4.0complete
baseline · v2.3.6archived
Evaluation · 240 attemptsIllustrative values
Success211
Failure18
Intervention11
Slip during transfer7
Double-layer pick6
Placement tolerance5
Robots online12 / 12
Tasks completed4 812
Interventions6
Attention1 cell

Infrastructure behind every deployment.

Deploying a robot is only the beginning. Monty provides the data, model, evaluation, and fleet infrastructure required to operate physical AI in production.

Console
Overview
Data
Models
Evaluations
Deployments
Fleet
Vial handling
Line 06
Vial Handling / Line 06Data
Session 3 102Task: vial stackSlip prevent412 demonstrations
A robot gripper holding a glass vial of blue liquid, with a live pressure map reading across the tactile fingertip pads
Tactile
Robot state
Action
1 284 / 1 412 reviewed
Metadata
OutcomeSuccess
QualityA — optimal friction
Task phaseLift → transfer
Annotations
GraspSlip detectedAdjust
Adaptation runsStatus
vial-stack-v2.4.1 · 3 102 demos · slip sensingTraining
vial-stack-v2.4.0 · 388 demosComplete
vial-stack-v2.3.6 · 388 demosArchived
vial-stack-v2.3.2 · 210 demosArchived
vial-stack-v2.4.1 · lossstep 12 400 / 20 000
Task configuration
Datasetvial-stack · tactile curated
Embodiment6-axis · two-finger
SensingOptoForce tactile pads
Checkpoints4 retained
Success · 211Failure · 18Intervention · 11Illustrative values
Success rate88%
Cycle time11.4 s
Intervention rate4.6%
Failure modes
Slip during transfer7
Double-layer pick6
Placement tolerance5
Evaluation set
Modelvial-stack-v2.4.1
Attempts240
ConditionsLine 04 · mixed fabric
CriteriaDefined per workflow
Robots online12 / 12
Need attention1
Tasks completed4 812
Success rate91%
Cycle time10.8 s
Deployed systemsStatus
Site A · Cell 01 · HandlingRunning
Site A · Cell 02 · InsertionRunning
Site B · Cell 01 · CappingIntervention
Site C · Cell 01 · Vial QCCommissioning
Interventions
Slip during transferSite B · Cell 01 · human fallback
Placement out of toleranceSite A · Cell 02 · review
Vial jamSite B · Cell 01 · resolved
Items to review6

Built by a team spanning the physical AI stack.

Our team spans AI post-training, robotics, mechanical engineering, embedded systems, data, and evaluation.

  • Stanford University
  • Carnegie Mellon University
  • UC Berkeley
  • UCLA
  • Meta
  • Amazon
  • Intel
  • Tesla
  • Sanctuary AI

Have a task you want to automate?

Monty systems are already live across customer and partner environments. Bring us the workflow and we’ll help determine the right path to deployment.