Skill UI

Guides engineers through DeepSpeed distributed training and NVMe I/O optimizations, covering ZeRO, pipeline parallelism, mixed precision, and DeepNVMe handles for synchronous and asynchronous tensor storage.

Orchestrates PyTorch-native TorchTitan 4D parallelism (FSDP2, TP, PP, CP) for large-scale LLM pretraining, scaling from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing to boot Llama 3.1, DeepSeek V3, or custom models.

This tool provides comprehensive Python solutions for processing batches of text-to-video requests using the Kling AI API. It handles controlled parallelism, rate-limiting, and detailed status tracking. Use this when building robust content pipelines that require generating multiple videos efficiently from a list of prompts, supporting synchronous, asynchronous, and webhook callback methods.

This skill provides comprehensive guidance on training Mixture of Experts (MoE) models, including architectures like Mixtral and DeepSeek-V3. It details techniques for achieving large-scale model capacity with significantly reduced computational costs (e.g., 5x reduction vs. dense models). Topics covered include top-k routing, load balancing, expert parallelism, and integration with frameworks like DeepSpeed and HuggingFace for efficient, resource-constrained training.

This guide details advanced architectural patterns for building complex AI systems using multiple Language Model instances (agents). It addresses the single-agent context bottleneck by distributing work, enabling parallelism, and facilitating specialization across different subtasks. Key patterns include Supervisor/Orchestrator, Peer-to-Peer/Swarm, and Hierarchical designs, providing guidelines for context isolation and effective coordination protocols.

Runs build, test, lint, serve, and other tasks in Nx workspaces via run, run-many, and affected flows, honoring the detected package manager and target metadata while exposing filters, parallelism, and caching controls for CI or local workflows.

Guides orchestrating multiple Task tool subagents with run_in_background for concurrent work, covering when to trigger parallelism, how to craft prompts, launch in one message, synchronize results, and integrate TodoWrite status updates.

Serve production-grade LLM APIs with vLLM, leveraging PagedAttention, continuous batching, quantization, and tensor parallelism to maximize throughput while keeping latency, GPU usage, and scalability under control for OpenAI-compatible deployments.

Provides production-ready Megatron-Core workflows for training 2B–462B parameter LLMs with tensor/pipeline/expert/context parallelism, maximizing GPU efficiency on H100 clusters.

Detects system CPU, GPU, memory, and disk availability before heavy scientific tasks, writing `.claude_resources.json` with strategic recommendations (parallelism, out-of-core, GPU acceleration) so analysts choose appropriate compute approaches.

Detects system CPU, GPU, memory, and disk availability before heavy scientific tasks, writing `.claude_resources.json` with strategic recommendations (parallelism, out-of-core, GPU acceleration) so analysts choose appropriate compute approaches.