Machine learning

Staff-level codebase health review.

Drive PostHog through the claude.ai `exec` connector — query analytics and entities (HogQL, insights, funnels, retention, web analytics), investigate LLM…

Elite AI Product Manager skill with expertise in AI product strategy, LLM product development, ML feature prioritization, AI ethics and fairness.

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

Use when the user asks about Claris MCP — Model Context Protocol integration with FileMaker (FM), enabling AI assistants to read data, run scripts, and…

Scan training data for benchmark contamination — n-gram overlap, embedding similarity, and exact canary detection against 60+ evaluation datasets.

BentoML — model serving and deployment.

Manages Volcano Engine ContextSearch control-plane and application-console operations.

Volcengine VOD audio and video processing tools skill.

Captum (PyTorch) — model interpretability and feature attribution.

Chroma — AI-native embedding database.

Clean, normalize, and prepare raw datasets for analysis or ML.

HuggingFace Diffusers library for diffusion models: text-to-image, image-to-image, inpainting, super-resolution.

Detect documentation-code drift.

Load, stream, process, and publish datasets with the HuggingFace datasets library.

Use this skill when you need evidence-bounded repeat inputs, version/model/prompt factors, invariants, variance evidence, and comparison boundaries; triggers…

Use this skill when you need evidence-bounded claim-to-source relations, unsupported assertions, abstention, uncertainty, and evidence review; triggers…

Use this skill when you need evidence-bounded grounding, relevance, completeness, citation support, abstention, and answer-level evidence in RAG outputs…

Use this skill when you need evidence-bounded query variants, chunking, filters, recall/precision proxies, ranking, freshness, and retrieval evidence…

결과를 숨긴 "블라인드" LLM 평가 패킷(사후 판단 편향 통제)은 두 방향으로 동시에 실패한다.

Cloud + LLM spend audit, right-sizing, unit economics.

Curate generated images into a clean training dataset, validate metadata, remove duplicates, and sync manifests/assets to Hugging Face.

给定 PMID / PMCID 取结构化全文（BioC XML/JSON）。封装 NLM BioC PubMed API（标题+摘要）和 BioC PMC OA API（开放获取全文），输出按章节切分的段落、表格上下文、RAG 友好 chunk。适合大批量入向量库。触发词：全文抽取 / BioC / RAG 入库…

Apply the DSG ONE / ProofGate truth boundary and claim policy.