Research and publications across neural architectures, applied AI, and adjacent ground.
Neural Architectures
Multilayer Perceptron and Kolmogorov–Arnold hybrid networks; differentiable simulation; partial neural warm deactivation; quantization at the limit.
Applied AI in Industry
Frontier model deployment in regulated, off-grid, and air-gapped contexts. Agentic workflows, RAG pipelines, edge inference, and fine-tuning for the places the cloud cannot reach.
Adjacent Ground
Hardware acceleration, real-time rendering, ML protocols, sparse-data forecasting, and the connective tissue that makes the rest of it ship.