Physical AI Platform Q2 Pilots Open

Give Your Robots Eyes That Think on the Edge

SensAI delivers onboard Vision AI with sub millimetre precision, running entirely on the robot, with no cloud round trip, no lag, and no data center dependency. Just fast, deterministic intelligence at the cell.

30ms
End to end latency
<1mm
Pose accuracy
0
Cloud dependencies
7
Robot brands supported

Integrated with leading robot brands & platforms

FANUC
ABB
KUKA
Universal Robots
Yaskawa
Comau
Flexiv
NVIDIA Isaac
FANUC
ABB
KUKA
Universal Robots
Yaskawa
Comau
Flexiv
NVIDIA Isaac
FANUC
ABB
KUKA
Universal Robots
Yaskawa
Comau
Flexiv
NVIDIA Isaac
What is SensAI

Vision AI That Lives on the Robot, Not the Cloud

SensAI is an onboard Vision AI platform by Spritle Software built for the next decade of industrial robotics. It delivers real time perception, pose estimation, and closed loop control, all running on an edge device mounted directly on or beside the robot cell, with zero cloud dependency.

Unlike traditional rule based machine vision or cloud dependent AI pipelines, SensAI uses foundation models running on NVIDIA Jetson hardware to handle novel parts, reflective materials, and high mix production lines out of the box, with sub 30ms end to end latency and sub millimetre precision.

Backed by 17 years of industrial automation engineering from Spritle's Chennai R&D Lab, SensAI is field proven across automotive, logistics, pharma, and aerospace, and ships with a customer grade SLA, not just a slide deck.

The Solution Packages

Three Power Packs. One Smarter Hospital.

  • Foundation Perception Custom VLM + SAM 2 + FoundationPose on Jetson Orin
    ACTIVE
  • Edge Inference TensorRT ROS 2 P99 latency <30ms
    ONLINE
  • Robot Integration EtherCAT RTDE 1kHz ROS2 control_msgs
    RUNNING
  • 6 DoF Pose Estimation Sub millimetre accuracy, frame by frame
    LIVE
  • Zero Shot Object Handling New SKUs in <5 minutes, no training required
    READY
Why Factories Are Stuck

The Infrastructure Hasn't Kept Up with Physical AI

Cloud Round Trip Latency

Typical cloud inference takes 180 to 400ms per frame, completely unusable for closed loop robot control. Any hiccup in connectivity stops the line.

Brittle Trained Once Vision

Legacy machine vision systems break when lighting shifts, new SKUs arrive, or production conditions change. They don't adapt, they fail, and re training takes weeks.

Massive Labelling Overhead

Traditional vision pipelines demand thousands of labeled samples per SKU. High mix, low volume lines never generate enough data to train reliably.

Teach Pendant Lock In

Rigid programmed trajectories break the moment a conveyor jitters, a pallet drifts, or a human introduces variance. Every deviation means downtime and re teach.

Cost of Unplanned Downtime

Every hour of unplanned downtime on an industrial line costs five figures or more. Brittle vision and cloud dependency are the silent root causes that never get tracked.

No SLA Backed AI Deployments

Most AI vendors deliver demos, not cells. There is no production grade commitment, no post deployment support, and no accountability when models degrade over time.

Applications

One Vision Stack. Six Battle Tested Applications.

01

Bin Picking

142 random, mixed parts per minute, including reflective metals, deformable items, and tangled geometries, gripped without reprogramming.

142 picks / min
02

Pick & Place

Sub-half-millimetre placement accuracy across varying surface tolerances, cycle after cycle. Handles everything from PCBs to curved housings.

<0.5mm accuracy
03

De-racking

Autonomous handling of body-in-white panels, painted parts, fragile glass, and sensitive composites in automotive paint shop environments.

Auto de-rack
04

Fastening

Torque-aware bolting, screwing, and riveting with 0.1mm sub-millimetre accuracy. ISO 5393 compliant on every cycle.

ISO 5393 compliant
05

Dispensing

Closed-loop adhesive bead tracking with ±0.1mm correction on complex 3D contours, even with varying flow rates.

±0.1mm bead control
06

Pharma Handling

GMP-validated cells for vial, blister, and ampoule handling under Annex 1 compliance. Every frame logged, every dispatch auditable.

GMP / Annex 1
Proven Results

Numbers That Matter on the Line.

Under 30ms End to End Latency

P99 budgeted at 30ms, deterministic, closed loop control that reacts faster than any cloud pipeline can even round trip.

Under 1mm Pose Accuracy

Sub millimetre 6 DoF pose estimation on reflective, transparent, and complex geometries, consistent across every cycle.

142 Picks Per Min Peak Rate

Achieved on a Tier 1 automotive stamping press feed with mixed, tangled, reflective metal parts, no reprogramming required.

18 to 32% Cycle Time Improvement

vs. stop and pick deployments using Dynamic Guidance on moving assembly lines.

Under 5 Min New SKU Onboarding

Zero training samples needed. Operators add new parts via a single reference photo. Foundation perception handles the rest.

17yrs Engineering Heritage

Spritle's Chennai R&D Lab brings 17 years of automation, vision-based inspection, and imitation learning heritage.

Architecture & Process

Industrial hardware meets foundation models.

Step 01

Scope & Site

Engineer-to-engineer briefing. We assess the cell, parts, and robot brand in a 30-minute session.

↳ Cell scoped & pilot defined
Step 02

Kit Configuration

SensAI Edge Kit configured for your robot brand, workspace geometry, and target application.

↳ Edge kit ready to ship
Step 03

Cell Integration

On-site installation and robot controller integration via EtherCAT, RTDE, or ROS 2 depending on brand.

↳ Robot connected & calibrated
Step 04

Validation & UAT

Live accuracy and throughput testing on your actual parts. Acceptance against agreed SLA metrics.

↳ SLA metrics verified
Step 05

Production Go Live

Full handover with observability dashboard, local audit logs, and Spritle R&D Lab on standby support.

↳ Cell live in production

Industries Served

AutomotiveLogisticsPharmaAerospaceElectronicsMetal Fab
Industries

Built for Industries Where Mistakes Are Expensive.

Automotive

Zero-escape QC on BIW lines. From sub-frame inspection to body-in-white de-racking.

Logistics

142 picks per minute on mixed, reflective, deformable bins. Zero-training perception.

Pharmaceutical

GMP-validated cells for vial, blister, and ampoule handling under Annex 1.

Aerospace

Sub-mm fastening and SOP verification on heavy parts. Torque-aware and traceable.

Why SensAI by Spritle

We Ship Cells. Not Slide Decks.

Most AI vendors stop at the demo. SensAI by Spritle Software brings 17 years of industrial automation engineering directly to your production floor.

Edge-First Architecture +
All inference runs on-robot. No cloud dependency, no network risk, deterministic latency guaranteed at P99.
Foundation Model Stack +
Custom VLM, SAM 2, and FoundationPose running on TensorRT, handling novel parts and complex geometries.
7 Robot Brands Supported +
FANUC, ABB, KUKA, Universal Robots, Yaskawa, Comau, and Flexiv, with native integration via EtherCAT, RTDE, and ROS 2.
SLA-Backed Deployment +
Customer-grade SLA on every cell shipped. Spritle R&D Lab provides ongoing support and model updates.
17 Years of Automation Heritage +
Spritle's industrial engineering background means we understand your line, constraints, and compliance.
30ms P99 end-to-end latency, on-robot
<1mm 6-DoF pose accuracy on any material
0 Cloud dependencies at runtime
17yr Industrial automation engineering heritage

Ecosystem Partners

NVIDIA IsaacNVIDIA HoloscanNVIDIA CosmosOmniverseRoboSuiteSprutCAM XFlexiv RDKROS 2TensorRT
From the Field

What Engineering Teams Say

We tried three vision vendors before SensAI. The difference is that Spritle's team actually understands the cell, not just the software. They scoped our stamping press feed in 30 minutes and had us running in under four weeks. We're at 142 picks per minute with zero reprogramming since go live.

Automation Manager Tier-1 Automotive Supplier, Chennai

The zero training perception was the biggest unlock for us. We run 200+ SKUs through the same cell. No labelled datasets, no retraining cycles, just a reference photo and it works. That was something none of our previous vendors could come close to delivering.

Robotics Lead Logistics & Warehousing, Singapore

Annex 1 compliance with full audit logs on every frame was non negotiable for us. SensAI delivered that out of the box, every dispatch is traceable, every frame logged to the on device NVMe buffer. Our QA team signed off in the first UAT session.

QA Engineering Lead Pharmaceutical Manufacturer, Singapore
FAQ

Frequently Asked Questions

Ship a Vision-Led Cell This Quarter

Ready to Put Intelligence on the Robot?

SensAI pilots are open now. Engineer-to-engineer, no NDAs to start, Singapore and Chennai teams on standby.

📞Book a 30-min Briefing

Technical briefing with SensAI engineers. Assess your cell, parts, and robot brand. No commitment required.

📦See the Edge Kit

Review the SensAI hardware and software stack, including Jetson Orin, 3D depth sensing, and foundation model runtime.

🚀Pilot This Quarter

Ship a live cell this quarter with measurable SLA-backed results from day one. Q2 slots are open now.

No NDAs to start Engineer-to-engineer Q2 pilots open SLA-backed deployment Zero cloud dependency