Experience

Electronic Warfare Associates

Signal analysis for models at the edge

Role
AI/Machine Learning Intern
When
Jun – Aug 2024
Where
Herndon, VA
Built with
Python, Keras, CNN variational autoencoders, spectrograms, IQ data
  • 600kIQ samples processed, 10M complex values each
  • 99,999spectrogram images, processed within edge-hardware limits

I architected spectrogram-based frequency-domain analysis for an enhanced edge model intended for military zones, processing a dataset of 99,999 images efficiently while staying within the target hardware’s constraints.

I engineered a custom CNN variational autoencoder in Keras to synthesize smooth signal manifolds, and processed a large-scale IQ dataset of 600,000 samples with 10 million complex values each.

NextTechnicaldle