Edge AI · On-device inspection
On-Device Edge Vision Defect Detection on ESP32-S3
A production conveyor inspection camera running a quantized INT8 CNN entirely on an ESP32-S3 — 18 FPS at 0.4 W, no cloud, 6× lower capex per station.
Edge AI & Perception
Edge AI & Machine Learning for Robotics & Hardware
YantriX engineers vision pipelines and edge machine learning systems directly coupled to physical robotics and industrial automation. From on-device YOLO models for robotic pick-and-place to automated defect inspection and Jetson inference pipelines integrated with ROS 2, we deliver deployable perception for hardware teams.
What we do
Practical support for targeted engineering work
We deliver applied machine learning and computer vision services focused on hardware and robotics. (1) Computer vision & perception — object detection, segmentation, optical quality inspection, and pose estimation. (2) Edge model optimization — converting models to TensorRT, ONNX Runtime, and INT8/FP16 quantization for low-latency inference on NVIDIA Jetson, Raspberry Pi, and microcontrollers. (3) Vision-guided robotics — integrating vision nodes with ROS 2 Nav2 and MoveIt for closed-loop manipulation and obstacle avoidance. (4) Sensor fusion & telemetry — processing multi-modal camera, LiDAR, and IMU data on embedded compute. (5) Technical search & engineering knowledge retrieval — structuring CAD and engineering documentation for searchable team access.
What problems we solve
- Deploy low-latency computer vision directly on edge hardware without depending on unstable cloud connectivity.
- Integrate camera perception pipelines directly with ROS 2 navigation and manipulator control loops.
- Optimize vision models for real-time inference on NVIDIA Jetson Orin Nano, AGX, and embedded GPUs.
- Automate visual quality inspection and defect classification on manufacturing assembly lines.
- Enable robotic arms and AGVs to reliably locate, identify, and handle industrial parts.
Tools we use
- Ultralytics YOLO (v8, v11), OpenCV, custom detectors
- TensorRT, ONNX Runtime, OpenVINO, TFLite / TFLite Micro
- NVIDIA Jetson (Nano, Orin Nano, Orin AGX), Raspberry Pi
- PyTorch, scikit-learn for model fine-tuning and evaluation
- ROS 2 (Humble / Jazzy) vision_opencv and image_transport
- Roboflow, Label Studio for industrial dataset labeling
- FastAPI, gRPC for local edge inference services
Deliverables
- Hardware-accelerated deployment binaries (TensorRT engines, ONNX models, quantized weights)
- ROS 2 perception nodes integrated with camera drivers and transform trees
- Trained and evaluated vision models with benchmark reports (FPS, mAP, latency, power draw)
- Edge deployment scripts, systemd service definitions, and automated health monitoring
- Complete engineering documentation, test datasets, and handover walk-throughs
Have a project requiring this engineering scope?
Share your technical requirements, 3D CAD files, or operating specs. NDA support is available where required before confidential file exchange.
Use cases
Industries where this service applies
We adapt the same engineering service to different product contexts depending on the load case, packaging problem, validation target, or deployment environment.
Manufacturing and industrial automation
Production-line fixturing, mechanical tooling, edge-inspection mounts, and machine-tending assemblies built to withstand factory environments.
Robotics and autonomous mobile systems (AMR / AGV)
Tailored engineering, mechanical design, and rapid prototyping solutions designed specifically for robotics and autonomous mobile systems (amr / agv) applications.
Automated optical inspection (AOI)
Tailored engineering, mechanical design, and rapid prototyping solutions designed specifically for automated optical inspection (aoi) applications.
Warehousing, sorting, and logistics
Tailored engineering, mechanical design, and rapid prototyping solutions designed specifically for warehousing, sorting, and logistics applications.
Precision agriculture and field robotics
Tailored engineering, mechanical design, and rapid prototyping solutions designed specifically for precision agriculture and field robotics applications.
Related work
Case studies tied to this service
See how this engineering capability applies across real project scopes, analysis goals, and physical prototype iterations.
Applied AI · Vision-guided robotics
Vision-Guided Robotic Bin-Picking Cell with 3D Pose Estimation
How a YOLOv11-Seg + 3D-pose stack on a Jetson Orin Nano replaced fixed-pose jigs in a 6-DOF robotic cell — sub-80 ms latency, 99.2% accuracy, 40% throughput gain.
GenAI · Retrieval-Augmented Generation
Engineering Retrieval System for CAD & Technical Documentation
How a hybrid-search RAG system over 40k engineering PDFs and CAD drawings cut average engineer-question turnaround from 35 minutes to 22 seconds, with grounded citations on every answer.
From the blog
Articles that support this service topic
Explore practical engineering guides related to design, manufacturing, simulation, and product development.
Applied AI
Deploying YOLOv11 to Jetson Orin Nano at 30 FPS
Walkthrough of shipping a segmentation-class YOLOv11 model to a Jetson Orin Nano at production latency — quantization, TensorRT conversion, and the pitfalls.
Applied AI
AI & ML Services in India: 2026 Buyer's Guide for Engineering Teams
A practical 2026 guide to sourcing AI and ML services in India — what's available, realistic project costs, when to outsource vs hire, and how to evaluate Indian AI vendors.
Applied AI
Computer Vision Development Services in India: Detection, Segmentation, OCR
A practical guide to computer vision development services in India — what use cases work, how vendors price, and what production deployment really looks like.
FAQ
Questions teams ask before they engage
Common questions teams ask before starting a project.
What edge AI and vision services do you offer?
We focus on applied computer vision and edge machine learning for physical hardware: object detection and classification (YOLO), edge acceleration (TensorRT/ONNX on Jetson), vision-guided robotics integrated with ROS 2, and optical defect inspection for production lines.
Can you optimize models to run on NVIDIA Jetson hardware?
Yes. We specialize in hardware-specific model optimization: FP16/INT8 quantization, TensorRT engine generation, memory footprint reduction, and thermal-aware benchmarking on Jetson Nano, Orin Nano, and Orin AGX.
How do your vision models integrate with robots?
We package perception models as modular ROS 2 nodes that publish detection bounding boxes, 3D poses, and point-cloud clusters directly into Nav2 costmaps or MoveIt planning scenes for real-time action.
Do you sign NDAs before discussing proprietary hardware or data?
Yes. Mutual NDA is available before confidential CAD, camera streams, or proprietary dataset exchange.
Where is the engineering team based?
Our engineering studio is in Surat, Gujarat, India. We work remotely with hardware teams across India and globally via milestone reviews and video walk-throughs.
Start your project
Deploying AI to Physical Hardware or Edge Devices? Build It to Ship
Share your target latency, hardware constraints (Jetson / ESP32), and data availability. We build models and firmware that run reliably in the field.