Systems Architecture Research In Machine Learning
CurrentConducting independent research and development in machine learning and AI, focusing on audio processing, natural language processing, and model interpretability.• Developed system for analyzing information flow in transformer architectures, revealing patterns in how different token types are processed through attention layers.• Architected experimental audio processing pipeline applying NLP architectural patterns to sound classification. Engineered complete system from raw audio processing through model training, including novel tokenization approach using K-means clustering on spectrograms.• Designed and implemented system for analyzing decision-making processes in Audio Spectrogram Transformers, including tools for isolating and interpreting critical segments of audio inputs.Live demos and detailed documentation available at demo.danavery.com.