Projects

Developing tools for building image and captioning datasets, training computer vision models, and building analysis tools.

Working on

• Building parameter efficient model analysis and tokenizer tools in the browser
• Building aesthetic prediction models (Preference, Rating)
• Enhancing training systems for Stable Diffusion
• Building Image captioning models using LoRA and other parameter efficient methods.
• Building dataset creation, filtering, captioning pipeline tools
• Building image rating and preference capturing tooling for aesthetic modeling (DPO, Aesthetic Predictor)
• Building stable diffusion scripts for analysis, conversion, pruning, scaling, evaluation
• Building transformers cross-attention visualization tooling with flexibility (DAAM)
• Experimenting with Rust candle ML for training, inference, and WASM
• Teaching and helping others learn to fine-tune foundation models