Discover Agent Skills for data science & ml. Browse 61skills for Claude, ChatGPT & Codex.
Identifies and resolves global consistency obstructions in topological AI systems using Čech cohomology and GF(3) balancing.
Models continuous performance curves and high-order musical gestures using topological category theory and Mazzola's Diamond Conjecture.
Navigates complex mathematical and ontological possibility spaces using Badiou-inspired event logic and Kripke semantics.
Evaluates research rigor and scientific claims by assessing methodology, statistical validity, and potential biases using standardized frameworks.
Orchestrates end-to-end MLOps pipelines from data preparation through production deployment and monitoring.
Architects sophisticated LLM applications using LangChain patterns for autonomous agents, stateful memory management, and modular chains.
Optimizes LLM performance and reliability through advanced prompting techniques like chain-of-thought and few-shot learning.
Builds Retrieval-Augmented Generation (RAG) systems to ground LLM applications with vector databases and semantic search capabilities.
Implements rigorous evaluation strategies for LLM applications using automated metrics, human-in-the-loop feedback, and advanced benchmarking.
Automates programmable chemical synthesis by treating chemical procedures as executable XDL code on modular robotic hardware.
Performs ultra-fast portfolio backtesting and trading strategy analysis using the Polars library.
Optimizes and manages PolicyEngine microsimulations with advanced caching, data access, and entity mapping patterns.
Performs objective technical analysis on weekly price charts to identify trends, support levels, and probabilistic price scenarios.
Generates policy impact simulations, distributional analyses, and interactive dashboards using the PolicyEngine framework.
Facilitates the development, simulation, and control of 3D-printed humanoid robots for reinforcement learning research.
Composes complex 3D environments and terrains for robotic simulation and reinforcement learning training.
Processes and analyzes complex physiological signals including ECG, EEG, and EDA for psychophysiology research and medical data science.
Implements a compositional AI framework based on category theory and GF(3) triadic balance for deterministic, self-modifying agent architectures.
Provides hardware specifications, MuJoCo MJCF models, and deployment workflows for the K-Scale flagship humanoid robot platform.
Performs advanced molecular analysis, descriptor calculation, and chemical informatics using the RDKit library.
Provides a unified framework for humanoid robot development, reinforcement learning training, and sim-to-real deployment.
Deploys trained Reinforcement Learning (RL) policies to real robots using high-performance Rust and ONNX Runtime.
Bridges the gap between robotic simulations and real-world deployment using maximum entropy reinforcement learning and information-theoretic alignment.
Optimizes trade execution using advanced algorithms like TWAP, VWAP, and Iceberg orders to minimize market impact and slippage.
Synthesizes Patrick Kenny's active inference framework with K-Scale's JAX/MuJoCo robotics stack for advanced predictive coding in robot locomotion.
Optimizes machine learning workflows on Databricks by implementing structured MLflow experiment tracking and model governance patterns.
Streamlines the creation, management, and serving of scalable feature stores within the Databricks MLOps ecosystem.
Standardizes robotics datasets and deploys edge-optimized vision-language-action models for embodied AI applications.
Deploys and manages production-grade machine learning models on Databricks with support for A/B testing and auto-scaling.
Optimizes AI agent memory usage and manages conversation context through advanced pruning and archiving strategies.
Scroll for more results...