DevOps & Infra

nota-america Bundle

Helps users find the right Azure RBAC role for an identity with least privilege access, then generate CLI commands and Bicep code to assign it. Also provides guidance on permissions required to grant roles. WHEN: bicep for role assignment, what role should I assign, least privilege role, RBAC role for, role to read blobs, role for managed identity, custom role definition, assign role to identity, what role do I need to grant access, permissions to assign roles.

nota-america Bundle

Query and analyze data in Azure Data Explorer (Kusto/ADX) using KQL for log analytics, telemetry, and time series analysis. WHEN: KQL queries, Kusto database queries, Azure Data Explorer, ADX clusters, log analytics, time series data, IoT telemetry, anomaly detection.

nota-america Skill

Package, deploy, and verify a playable Three.js or web game. Use for release builds, asset delivery, private/public deployment, production smoke tests, browser proof, release notes, rollback readiness, and cleanup of temporary QA resources.

nota-america Bundle

Track ML experiments, manage model registry with versioning, deploy models to production, and reproduce experiments with MLflow - framework-agnostic ML lifecycle platform

nota-america Bundle

Managed vector database for production AI applications. Fully managed, auto-scaling, with hybrid search (dense + sparse), metadata filtering, and namespaces. Low latency (<100ms p95). Use for production RAG, recommendation systems, or semantic search at scale. Best for serverless, managed infrastructure.

adu2021 Skill

Train multi-agent reasoning systems with decoupled reward signals and pipeline parallelism—enable specialized Solver/Verifier/Corrector agents to iteratively refine solutions without waiting for full trajectories, handling extended reasoning up to 320K tokens.

adu2021 Skill

Improve language model performance on complex numerical reasoning over tables through three-stage pipeline: decompose questions into sub-questions, sanitize and clean table data, then generate executable Python code for precise computation.

adu2021 Skill

Evolutionary architecture search discovering optimal depth-width ratios and operator combinations under deployment latency constraints using augmented scaling laws. Deploy when you need fastest inference per latency target with mixed attention mechanisms.

adu2021 Skill

Uses distilled language model (DLM) for KV cache token retrieval, reducing layer-wise retrieval parameters by 90% via head-level attention weights, with asynchronous prefetching and elastic CPU offloading. Deploy for ultra-long-context inference with memory constraints.

adu2021 Skill

Replace Docker environments with learned LLM surrogates comprising a lightweight sandbox for deterministic file operations, a transition model predicting execution feedback, and a reward model acting as virtual test runner. Reduces training infrastructure costs while scaling SWE agent training data.

adu2021 Skill

Cascaded VAE+SigLIP encoders creating single continuous representation space supporting both vision understanding and generation, trained jointly on both tasks without format mismatches. Deploy for unified multimodal models where understanding and generation enhance each other.

adu2021 Skill

Implement CASTLE, a causal attention mechanism that dynamically updates key representations as context expands. Reduces validation loss by 0.006-0.037 across model scales while maintaining O(L²d) training complexity and O(td) decoding speed. Deploy for improved language model perplexity without inference overhead.

adu2021 Skill

Train vision-language models on hour-long videos using a two-stage pipeline combining supervised fine-tuning with GRPO, reaching 71% accuracy on VideoMME while supporting 8,192 frames through efficient multi-modal sequence parallelism.

adu2021 Skill

Deploy frontier-level reasoning with only 11B active parameters using sparse MoE with 288 routed experts plus shared expert. Use Metropolis Independence Sampling-Filtered Policy Optimization (MIS-PO) to stabilize RL training at scale, replacing continuous importance weighting with discrete filtering that ensures trust region stability.

adu2021 Skill

Extend visual reasoning to spatial-temporal sequences via two-agent reasoning+summary pipeline with ST-GRPO (temporal alignment) and J-GRPO (evaluative robustness) algorithms. Achieves +8.1% on image and +6.9% on video benchmarks through autonomous data generation and self-evolving rewards; enables continuous improvement loops for visual reasoning without human annotation.

adu2021 Skill

Generates high-quality audio through three-stage CoT reasoning pipeline: foundational foley synthesis, object-focused refinement, and instruction-guided editing. Uses fine-tuned VideoLLaMA for reasoning and flow-matching audio foundation model. Apply for professional audio design workflows or video-to-audio applications.

adu2021 Skill

Train multimodal agents to dynamically invoke tools (code execution, web search) within reasoning loops through a two-stage pipeline combining cold-start supervised learning with reinforcement learning—enabling task-adaptive tool invocation for perception, reasoning, and retrieval tasks.

adu2021 Skill

Improve mathematical problem-solving in multimodal models by decoupling visual perception from inference reasoning. A two-stage pipeline extracts essential visual information and reasoned properties before passing enriched text to inference models, dramatically improving accuracy on visual math problems.

adu2021 Skill

Systematically train 8B web agents to solve complex information-seeking tasks through model-based exploration and long-to-short query evolution, achieving state-of-the-art performance via SFT+RL training pipeline with progressive context expansion to 128K tokens and 100 tool turns.

adu2021 Skill

Segment novel 3D point cloud classes with few support samples by combining dense but noisy pseudo-labels from 3D vision-language models with precise sparse few-shot annotations. GFS-VL adapts to new classes while retaining base class performance, using prototype-guided filtering and adaptive infilling strategies ideal for applications with limited labeled training data.

adu2021 Skill

Mechanistic Interpretability (MI) has emerged as a vital approach to demystify the opaque decision-making of Large Language Models (LLMs). However, existing reviews primarily treat MI as an observational science, summarizing analytical insights while lacking a systematic framework for actionable intervention. To bridge this gap, we present a practical survey structured around the pipeline: 'Locate, Steer, and Improve.' We formally categorize Localizing (diagnosis) and Steering (intervention) met...

adu2021 Skill

Deploy compact vision-language-action models that run on consumer GPUs for natural language robot control.

adu2021 Skill

Build multilingual LLMs that excel at machine translation while preserving general-purpose capabilities through a staged training pipeline combining continued pretraining, SFT, preference optimization, and RL with verifiable rewards.

adu2021 Skill

Build open-source agents for document question-answering by modeling DocQA as information-seeking with explicit tool utilization. DocDancer uses an exploration-then-synthesis pipeline to generate high-quality training data, addressing the scarcity that limits agent-based document understanding systems.

adu2021 Skill

Combines SFT and DPO with robust data selection pipeline using multidimensional quality metrics. Achieves DeepSeek-R1 performance with 12% training data, enabling efficient reasoning model alignment.

nota-america Bundle

Scores inbound HubSpot leads by engagement signals, company fit, and urgency markers to produce a "call these 5 today" list with talking points, drafts the follow-ups, and blocks Calendar time. Use when the user asks to prioritize leads, who to call first, or about their pipeline.

nota-america Bundle

Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.

nota-america Skill

Generates a one-page Monday morning briefing — cash, sales, pipeline, week ahead, top three to-dos. Accepts optional post destination and save-to arguments.

nota-america Bundle

Guidance for instrumenting webapps with Azure Application Insights. Provides telemetry patterns, SDK setup, and configuration references. WHEN: how to instrument app, App Insights SDK, telemetry patterns, what is App Insights, Application Insights guidance, instrumentation examples, APM best practices.

nota-america Bundle

Produces a one-page cross-functional business snapshot for SMB owners — cash position (QuickBooks), sales trend (PayPal/Square), pipeline movement (HubSpot), this week's commitments (Calendar), urgent watch-list items (Gmail/Slack), and the single most important thing needing attention today. Proactively tries every available connector and gracefully scopes to whatever is connected — one connector gives a partial pulse; the full stack gives the full picture. Trigger when the user asks how the business is doing, wants a snapshot, a weekly summary, a Monday brief, or says anything like "what am I missing" or "catch me up on the business."

nota-america Skill

Track and manage recruiting pipeline stages. Trigger with "recruiting update", "candidate pipeline", "how many candidates", "hiring status", or when the user discusses sourcing, screening, interviewing, or extending offers.

undermybelt Skill

AutoR — human-centered 研究执行框架。8阶段 pipeline，产出可复现 artifact。

undermybelt Bundle

Comprehensive OSINT methodology for external red-team operations and authorized attack-surface assessments. Covers the 5-stage recon pipeline, asset-graph discipline, severity rubric, confidence upgrade workflows, time budgeting, identity-fabric mapping, breach×identity correlation, detectability tagging, detection-aware probing, WAF/CDN bypass, vulnerability prioritization, phishing infrastructure planning, bug bounty submission, and client deliverable templates. Use when planning or executing reconnaissance against authorized targets, mapping an organization's external attack surface, investigating a person/entity, or producing client deliverables.

undermybelt Skill

RFC-driven multi-agent DAG execution pattern with quality gates, merge queues, and work unit orchestration.

ferroxlabs Bundle

Advanced model evaluation covering LLM benchmarks, evaluation frameworks (lm-evaluate-harness, HELM, RAGAS), leaderboard interpretation, custom metrics design, human evaluation protocols, automated LLM-as-judge patterns, and evaluation pipeline architecture for both traditional ML and generative AI systems. Use when the user asks about model evaluation specialist, model evaluation specialist best practices, or needs guidance on model evaluation specialist implementation. Do NOT use when the user needs a different specialized skill or is asking about an unrelated technology domain.

ferroxlabs Bundle

Designs a streaming data architecture defining event schema, producer configuration, consumer processing logic, state management approach, and fault tolerance strategy. Produces a system specification document for real-time data pipelines. Use when the user asks to design a real-time data pipeline, build an event-driven architecture, implement stream processing, or plan a system that processes data as it arrives rather than in batches. Do NOT use for batch ETL pipelines (use etl-pipeline-design), data warehouse design (use data-warehouse-design), or API data extraction (use api-data-extraction).