Building the whole path around the model.
My work sits between machine learning, product engineering and operations. I am interested in systems where context must be retrieved, multiple tools or agents must coordinate, and a person still needs a clear point of control.
That perspective extends beyond individual models. Useful AI depends on legible processes, decision history, good datasets, deployment infrastructure and evaluation that reflects the actual job being done.
Research and teaching as engineering tools.
My research on diffusion-model hallucinations used complexity–entropy analysis, Rényi entropy, Bandt–Pompe representations and clustering. This site also turns handwritten NLP and generative-models material into structured, permanent technical notes.
When deciding between complexity and simplicity, choose simplicity. The important asset is the content.