Skip to content
View kqb's full-sized avatar
🎯
Focusing
🎯
Focusing
  • Dallas, TX

Block or report kqb

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
kqb/README.md

Katie (kqb)

Senior software engineer, 12 years production experience. Currently focused on multi-agent systems and LLM inference optimization.

Current work

  • mlx-od-moe — On-Demand Mixture of Experts for Apple Silicon. Run 375GB models in 192GB RAM via memory-mapped expert loading.
  • AgentOS — Multi-agent orchestration layer for Windsurf IDE over Chrome DevTools Protocol.
  • cascade-multiagent — Programmatic multi-agent control for Windsurf Cascade via CDP.
  • live-translation-local — Real-time multi-modal conversation system. Whisper, NLLB-200, pyannote.audio, Even Realities G2 glasses.
  • localllm-hub — Local LLM routing and serving layer.

Focus areas

Inference optimization · Agent orchestration · Local LLMs · Real-time pipelines · Apple Silicon · Multi-agent systems

Based in Dallas, TX.

Pinned Loading

  1. mlx-od-moe mlx-od-moe Public

    On-Demand Mixture of Experts for Apple Silicon — run 375GB models in 192GB RAM

    Python 4 1

  2. live-translation-local live-translation-local Public

    Multi-modal conversation intelligence system: Real-time transcription, translation, speaker recognition, AR glasses output, and semantic memory capture. Integrates Whisper, NLLB-200, pyannote.audio…

    Python 2

  3. agent-orchestra agent-orchestra Public

    Multi-agent orchestration framework with structured communication protocols, event-driven coordination, and task dependency management for AI coding agents.

  4. AgentOS AgentOS Public

    AgentOS is an extended agent orchestration layer that operates within Windsurf IDE. It provides multi-agent coordination capabilities through Chrome DevTools Protocol integration.

    TypeScript

  5. cascade-multiagent cascade-multiagent Public

    Programmatic multi-agent control for Windsurf Cascade via CDP

    JavaScript

  6. localllm-hub localllm-hub Public

    Local LLM routing and serving layer

    JavaScript