I build systems that make models faster and smaller — from pruning multi-task networks to accelerating image/ video diffusion transformers.

• Efficient Generative Models — sparse attention, linear attention, and inference acceleration for image/video diffusion transformers
• Multi-Task Learning — adaptive pruning frameworks for jointly-trained models
• Model Compression — structured pruning, quantization, knowledge distillation

I run a fleet of 8 specialized AI agents on OpenClaw that orchestrate Claude, Codex, and Gemini CLI to handle my daily workflow autonomously.

They dispatch coding tasks to remote GPU clusters, monitor my inbox, write research reports, review PRs, and sync everything to Notion — while I sleep.

The best AI setup isn't the smartest model — it's the best memory, tools, and orchestration around it.