I will deploy your ml models


Über diesen Service
Get your ML model running fast on edge hardware.
I convert PyTorch or TensorFlow models to ONNX, then optimise for NVIDIA Jetson (TensorRT), Intel edge (OpenVINO), or Raspberry Pi. Typical result: 3x to 10x faster inference on the same device.
What you get:
- Optimised model file (TensorRT engine, OpenVINO IR, or ONNX)
- Benchmark on your target hardware (FPS, latency, memory)
- Accuracy comparison against your original model
- Written notes explaining the trade-offs I made
I have 7+ years of production computer vision experience, deployed at industrial customers across Europe. I am currently CTO of an industrial AI startup.
Send me: your model file (or a Colab notebook), target hardware, and target FPS. I will confirm feasibility before you order.
Hardware I regularly work with:
- NVIDIA Jetson Nano, Xavier NX, Orin Nano, Orin AGX
- Raspberry Pi 4 and 5 (with or without Coral USB accelerator)
- Intel NUC, Intel edge devices via OpenVINO
Models I have optimised: YOLO family (v5, v7, v8), Mask R-CNN, U-Net, EfficientNet, MobileNet, custom CNNs.
Lerne Mutte U kennen
CTO Edge AI Engineer I deploy computer vision on Hardware
- AusFrankreich
- Mitglied seitAug. 2025
Sprachen
Englisch, Französisch
Meine weiteren Dienstleistungen im Bereich KI-Entwicklung
FAQ
What model formats can you convert?
PyTorch (.pt / .pth), TensorFlow (SavedModel / .h5), ONNX, Keras.
Which target hardware do you support?
NVIDIA Jetson (all generations), Intel via OpenVINO, Raspberry Pi 4/5, Coral USB, x86 CPUs.
Will accuracy drop after quantization?
Usually 0.5% to 2% for FP16, 1% to 5% for INT8 depending on the model. I run a full comparison before delivery.
Can you also help with training the original model?
Yes, get the Premium tier or contact me for a custom offer.

