Skill UI

High-performance RLHF framework that combines Ray, vLLM, and ZeRO-3 to accelerate PPO/GRPO/RLOO/DPO training for 7B-70B+ models on distributed GPU clusters, providing a single-stack workflow for reward models and policy optimization.

Ray Data offers scalable distributed data processing for ML workloads, streaming across CPU/GPU clusters and integrating with PyTorch, TensorFlow, and Ray Train. Use it for batch inference, ETL, and multi-modal preprocessing from a laptop up to hundreds of nodes.

Ray Train orchestrates distributed PyTorch, TensorFlow, and HuggingFace workloads from laptops to clusters, automating GPU placement, fault tolerance, checkpointing, and hyperparameter sweeps so teams can scale massive model training with minimal code changes.

Guides training large language models with Volcano Engine's verl library, covering PPO/GRPO workflows, distributed backends, multi-turn rollouts, reward configs, critic-enabled PPO, and production deployments using Ray workers.

A comprehensive, systematic framework for auditing and improving web UI designs and frontend code. It teaches core principles like grayscale-first design, constrained scaling (for spacing, type, and color), and establishing clear visual hierarchy. Use this guide when needing to refine an existing UI that 'looks off,' or requires professional structural and aesthetic improvements.

Implement Roblox physical simulation and queries with assemblies, anchoring, constraints, collision groups, raycasts, overlap queries, impulses, forces, network ownership, and physics cleanup for collisions, hit detection, projectiles, vehicles, and knockback.

Distributed data processing library for ML workloads with streaming execution across CPU/GPU, supporting Parquet, CSV, JSON, images, batch inference, ETL, and Ray Train/PyTorch/TensorFlow integration.

Ray Train enables distributed machine learning training orchestration across clusters, scaling frameworks like PyTorch, TensorFlow, and HuggingFace from single laptops to thousands of nodes. It includes built-in hyperparameter tuning with Ray Tune, fault tolerance, and elastic scaling capabilities. Ideal for training massive models across multiple machines or running distributed hyperparameter sweeps efficiently.