Browse categories
Explore
Fiverr Pro
English
$
USD
I will build an edge ai computer vision app for raspberry pi or jetson
Ai and Computer vision Solutions
I'm an AI Engineer specializing in Computer Vision, Machine Learning, and intelligent document processing. If you need a custom-trained model, a real-time detection pipeline, or an OCR-powered automat...
About this Gig
Tired of slow, expensive cloud AI ? Run it on your device instead.
I build and deploy computer vision systems that run directly on Raspberry Pi and Jetson hardware real-time, offline, no cloud costs, no latency. Detection, tracking, counting, anomaly detection all running locally on your edge device.
What makes this different from generic CV gigs:
Most sellers hand you a Python script that runs on a laptop. I optimize specifically for edge hardware ONNX export, TensorRT acceleration, INT8/FP16 quantization so your model actually runs at usable FPS on constrained hardware.
What I work with:
- Hardware: Raspberry Pi 4/5, Jetson Nano, Jetson Orin, Jetson Xavier
- Models: YOLOv8, MobileNetV3, EfficientDet, custom architectures
- Optimization: ONNX, TensorRT, OpenVINO, model quantization
- Languages & tools: Python, OpenCV, PyTorch, FastAPI
Typical use cases:
- Security cameras, industrial defect detection, retail people counting, agricultural monitoring, robotics vision.
What you get:
- Fully working model running on your device
- Optimized for real FPS on your specific hardware
- Clean documented Python code
- Setup and installation guide
- Source code always included, no exceptions
My Portfolio
FAQ
Do I need to send you my physical device?
No. I develop and optimize everything on equivalent hardware on my end and deliver ready-to-run code with a full setup guide. You just run it on your device.
What FPS can I expect on Raspberry Pi?
Depends on model and resolution. A quantized YOLOv8n with TensorRT typically hits 8–15 FPS on RPi 5 and 25–40 FPS on Jetson Nano. I'll tell you realistic numbers for your exact use case before we start.
I have no dataset. Can you still build this?
Yes. For common objects (people, vehicles, faces) pretrained weights work well out of the box. For niche use cases like industrial parts or specific crops, you'll need at least 200–300 labeled images per class I can help with annotation as an add-on.
What if my device isn't Raspberry Pi or Jetson?
Message me with your device specs before ordering. I've worked with other ARM-based and embedded hardware and can usually make it work just need to confirm first.
Will the model run without internet?
Yes. That's the whole point of edge AI. Everything runs locally on your device with zero cloud dependency once deployed.
Can you add a FastAPI server so my app can call the model?
Yes. This is included in the Premium package and available as an add-on ($70) for Basic and Standard.
