Project Description
I am bringing together three work-streams—hardware, machine-learning and front-end—to create a compact Near-Infrared (NIR) analysis unit based on a Raspberry Pi.
1) NIR Processing PCB
I already have a Pi-based prototype on a breadboard; now I need a production-ready PCB that hosts the NIR sensor, analogue front end, ADC, power management and the Pi connector. You will provide the KiCad (or Altium) schematic, layout, BOM and Gerber files, then iterate with me until the board is ready for fabrication and assembly.
2) AI Training
Once raw spectra reach the Pi, I want an end-to-end training pipeline that turns them into actionable insights. Expect to clean the data, design a model in Python (TensorFlow/PyTorch), and train it so future samples are processed locally on the Pi with acceptable inference speed. I will supply the initial dataset and help define the target outputs; you will handle feature extraction, model architecture and on-device optimisation.
3) GUI Design
The Pi runs a small touchscreen. I need a clear, responsive GUI—written in PyQt, Tkinter or another lightweight framework—that displays live spectra, inference results and logging controls. Smooth user interaction is essential; visual polish is a bonus.
Deliverables
• PCB design package (schematic, layout, Gerbers, BOM)
• Pi firmware to stream sensor data
• Reproducible AI training scripts + trained model file
• Stand-alone GUI application with install instructions
• Short build/run guide so I can replicate everything on my side
I am comfortable breaking the work into milestones if that helps us track progress. Let me know which part you would like to start with first and any clarifying questions you have—I’m ready to get moving.