Skill UI

A comprehensive collection of battle-tested, structured prompts designed to elevate interactions with Large Language Models (LLMs). Covers advanced use cases including expert coding reviews, system architecture design, performance analysis, technical writing, and creative brainstorming, ensuring high-quality, actionable AI output.

This skill demonstrates the multi-model pattern for creating robust and clean API contracts using Pydantic. It guides the creation of separate models for creation requests (Create), updates (Update), responses (Response), and database documents (InDB), ensuring type safety and clear data validation for Python backends. It also covers best practices like handling camelCase aliases and optional fields.

Guides adding PyTorch FSDP2’s fully_shard integration to training loops, covering correct distributed init, bottom-up sharding, mixed precision/offload policies, DTensor-aware optimizers, and distributed checkpointing so oversized models or DeviceMesh sharding goals work reliably.

Automates PyTorch training workflows by guiding data preparation, model configuration, hyperparameter tuning, and experiment tracking while following industry best practices for production readiness.

Pyvene is a declarative framework designed for performing causal interventions on PyTorch neural networks. It allows researchers to conduct advanced experiments such as activation patching, causal tracing (ROME-style), and interchange intervention training (IIT). Use this library when you need to test causal hypotheses about model behavior, deeply interpret model components, or ensure reproducibility in advanced AI research.

This skill simulates the role of a quantitative analyst, providing comprehensive tools for building complex financial models, backtesting trading strategies, and analyzing market data. It covers advanced topics like portfolio optimization (Markowitz, Black-Litterman), calculating risk metrics (VaR, Sharpe ratio), and performing statistical arbitrage. Ideal for quantitative finance professionals needing rigorous data analysis, robust backtesting, and risk-adjusted performance evaluation.

Quantizes HuggingFace LLMs to 8-bit or 4-bit with bitsandbytes, cutting memory by up to 75% while keeping accuracy, and supports QLoRA fine-tuning plus 8-bit optimizers for faster, memory-efficient training.

Expert in designing and implementing advanced Retrieval-Augmented Generation (RAG) pipelines. This role focuses on optimizing the entire information retrieval process—from document chunking and embedding model selection to vector database architectures and hybrid search strategies. The goal is to bridge the gap between raw knowledge documents and accurate LLM understanding, drastically reducing hallucination and enhancing context relevance for enterprise applications.

This comprehensive workflow guides the end-to-end implementation of Retrieval-Augmented Generation (RAG) systems. It covers critical stages, including selecting optimal embedding models, setting up vector databases, designing chunking strategies, implementing hybrid retrieval, integrating Large Language Models (LLMs), and performing rigorous evaluation. Use this when building knowledge-grounded AI applications or advanced semantic search systems.

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.

This guide provides comprehensive best practices for creating high-quality videos using Remotion within a React environment. It covers advanced techniques for dynamic content generation, including integrating 3D models (Three.js), managing diverse assets like audio, video, and fonts, implementing complex animations, data visualization, and generating precise captions. Use this skill when developing professional, robust video generation workflows in React.

Automate Replicate AI workflows via the Composio MCP integration to inspect model schemas, run predictions (sync or async), upload input files, list versions, and manage prediction history and related assets.