Skill UI

Adds PyTorch FSDP2 setup to training scripts so you can initialize torchrun, bottom-up shard modules, configure mixed precision/offload, and checkpoint via DCP when models outgrow a single GPU.

Ray Data is a scalable distributed data processing library for ML workloads, streaming execution across CPU/GPU, integrates with Ray Train, PyTorch, TensorFlow, and handles Parquet/CSV/JSON/images for preprocessing, ETL, or batch inference from single node to cluster.

Tracks CPU, memory, disk/network I/O, DB connections, and threads via the resource-usage-tracker plugin to reveal performance bottlenecks, recommend right-sizing, and guide cost-saving actions.

Guides production-ready patterns for Unity’s DOTS stack (ECS, Job System, Burst) to build data-oriented, CPU-efficient games, manage large entity counts, and replace OOP logic with parallelized, high-performance workflows.

Assesses available CPU, GPU, memory, and disk before computationally intensive scientific work, emitting a JSON summary with strategic recommendations on parallelism, memory handling, and GPU backends so you can choose the optimal compute pattern.

Performance Profiling Toolkit systematically analyzes Node.js, Python, and Go applications to uncover CPU, memory, bundle, and database bottlenecks, then validates fixes with flamegraphs, bundle reports, and k6/Artillery load runs.

Diagnose Claude Code and local system slowdowns by measuring CPU, RAM, disk, browser, and network health, flagging bottlenecks, and recommending corrective actions or background monitoring.