Hardware accelerator design for neural network inference

This Git repo hosts the material and documentation for a course I delivered on AI inference acceleration.

The final deliverable is a pre-trained neural network inference engine hosted on an FPGA. The neural network is a digit classifier using data from the MNIST set.

While this course is adapted for a classroom setting, you are free to view the course handouts.

Housekeeping

The project is to be taught over 12 contact hour sessions with around 3-4 hours of independent study per session. In total, budget around 70 hours.

It targets curious young minds with some background in Python, C and computing basics.

License

MIT

Copyright

Matias Wang Silva 2025