Machine learning

Run bounded LLM-led Kinic Wiki maintenance through inspection, evidence-backed edits, and reinspection.

Captures screenshots, PDFs, and visual artifacts from pages driven by cdp scripts or the interactive CLI.

Skill instructions for creating interactive data-table artifacts with sortable, filterable tabular data, safe JSON embedding, theming, and…

Local LLM integration and management

从无 embedding 的 PyTorch/ safetensors 小型语言模型中恢复 KCTF 第九题 flag；将 argmax 约束转化为整数线性规划并验证输出。

Build and navigate DSP (Data Structure Protocol) — graph-based long-term structural memory of codebases for LLM agents.

Use a live Jupyter kernel for stateful, iterative Python execution via hamelnb.

Manter o pipeline de treino contínuo de IA em `E:/GitHub/aidiabetic-research/src/` sempre: 1.

Cost optimization patterns for LLM API usage — model routing by task complexity, budget tracking, retry logic, and prompt caching.

Cite file:line, command output, or a captured artifact for every claim; treat should/presumably/likely as unfinished work.

Implement and review Braintrust SDK instrumentation for LLM providers and agent frameworks.

Create or edit a vllm-on-tap custom preset in .vot/presets/ from a free-form request - a model of interest, a Hugging Face link, a GPU, a constraint.

Generate The Architect's Digest — a daily curated briefing of the most relevant stories in architecture, distributed systems, cloud, languages, tooling, AI…

Coordinate multiple coding agents in shared repos with task packets, worktree or branch isolation, path ownership, verification loops, and PR or CI…

Production machine-learning engineering workflow for data contracts, reproducible training, model evaluation, deployment, monitoring, and rollback.

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", "tear this apart", reviewing…

Call Gala from Python — PyO3 bridge, Jupyter kernel, PyTorch integration, NumPy arrays

Vercel AI SDK expert guidance.

Implement or debug ChatWaifu realtime media, Pipecat, transport, VAD, STT, turn detection, LLM/TTS streaming, playback, interruption, cancellation…

Use only when creating new registrable ML components that require Factory or Registry patterns.

Verifies semantic and behavioral correctness of LLM, retrieval, routing, memory, fallback, and tool-using agent systems.