Hugging Face Transformers best practices including model loading, tokenization, fine-tuning workflows, and inference optimization. Use when working with transformer models, fine-tuning LLMs, implementing NLP tasks, or optimizing transformer inference.
Use when building production-grade GenAI, Agentic Systems, Advanced RAG, or setting up rigorous Evaluation pipelines.
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.
Implement comprehensive evaluation strategies for LLM applications using automated metrics, human feedback, and benchmarking. Use when testing LLM performance, measuring AI application quality, or establishing evaluation frameworks.
State-of-the-art Bayesian optimization for DSPy programs with joint instruction and demo tuning
Fine-tune LLM weights using DSPy's BootstrapFinetune optimizer
Forecast categories, weighted pipeline calculations, deal scoring models, and forecast accuracy metrics.
Build LLM-powered applications and pipelines using proven methodology - task-model fit analysis, pipeline architecture, structured outputs, file-based state, and cost estimation. Use when building AI features, data processing pipelines, agents, or any LLM-integrated system. Inspired by Karpathy's methodology and production case studies.
MLOps best practices for model versioning, experiment tracking, deployment, monitoring, and retraining workflows. Covers reproducibility, CI/CD for ML, model registry, and production ML system design.
Model evaluation metrics, testing protocols, and performance assessment for Somali dialect classification. Covers accuracy, F1-score, confusion matrix analysis, per-dialect performance, and evaluation best practices for multi-class classification tasks.
This skill provides comprehensive guidance for using the Replicate CLI to run AI models, create predictions, manage deployments, and fine-tune models. Use this skill when the user wants to interact with Replicate's AI model platform via command line, including running image generation models, language models, or any ML model hosted on Replicate. This skill should be used when users ask about running models on Replicate, creating predictions, managing deployments, fine-tuning models, or working with the Replicate API through the CLI.
SAP HANA Machine Learning Python Client (hana-ml) development skill. Use when: Building ML solutions with SAP HANA's in-database machine learning using Python hana-ml library for PAL/APL algorithms, DataFrame operations, AutoML, model persistence, and visualization. Keywords: hana-ml, SAP HANA, machine learning, PAL, APL, predictive analytics, HANA DataFrame, ConnectionContext, classification, regression, clustering, time series, ARIMA, gradient boosting, AutoML, SHAP, model storage
Track and optimize agent specialization during methodology development. Use when agent specialization emerges (generic agents show >5x performance gap), multi-experiment comparison needed, or methodology transferability analysis required. Captures agent set evolution (Aₙ tracking), meta-agent evolution (Mₙ tracking), specialization decisions (when/why to create specialized agents), and reusability assessment (universal vs domain-specific vs task-specific). Enables systematic cross-experiment learning and optimized M₀ evolution. 2-3 hours overhead per experiment.
Validate methodology effectiveness using historical data without live deployment. Use when rich historical data exists (100+ instances), methodology targets observable patterns (error prevention, test strategy, performance optimization), pattern matching is feasible with clear detection rules, and live deployment has high friction (CI/CD integration effort, user study time, deployment risk). Enables 40-60% time reduction vs prospective validation, 60-80% cost reduction. Confidence calculation model provides statistical rigor. Validated in error recovery (1,336 errors, 23.7% prevention, 0.79 confidence).
Achieve 3-4 iteration methodology convergence (vs standard 5-7) when clear baseline metrics exist, domain scope is focused, and direct validation is possible. Use when you have V_meta baseline ≥0.40, quantifiable success criteria, retrospective validation data, and generic agents are sufficient. Enables 40-60% time reduction (10-15 hours vs 20-30 hours) without sacrificing quality. Prediction model helps estimate iteration count during experiment planning. Validated in error recovery (3 iterations, 10 hours, V_instance=0.83, V_meta=0.85).
Cross-model performance optimization and scaling configurations for autonomous agents
Universal model detection and capability assessment for optimal cross-model compatibility
Advanced contextual pattern recognition with project fingerprinting, semantic similarity analysis, and cross-domain pattern matching for enhanced learning capabilities
Multi-layer validation strategy for CFN Loop to prevent "consensus on vapor" and ensure robust, high-quality deliverables
Implement ReasoningBank adaptive learning with AgentDBs 150x faster vector database. Includes trajectory tracking, verdict judgment, memory distillation, and pattern recognition. Use when building self-learning agents, optimizing decision-making, or implementing experience replay systems.
Expert prompt optimization for LLMs and AI systems. Use PROACTIVELY when building AI features, improving agent performance, or crafting system prompts. Masters prompt patterns and techniques.
Expert skill for integrating local Large Language Models using llama.cpp and Ollama. Covers secure model loading, inference optimization, prompt handling, and protection against LLM-specific vulnerabilities including prompt injection, model theft, and denial of service attacks.
Automatically intercepts and optimizes prompts using the prompt-learning MCP server. Learns from performance over time via embedding-indexed history. Uses APE, OPRO, DSPy patterns. Activate on "optimize prompt", "improve this prompt", "prompt engineering", or ANY complex task request. Requires prompt-learning MCP server. NOT for simple questions (just answer them), NOT for direct commands (just execute them), NOT for conversational responses (no optimization needed).
Create AI learning plugins using AgentDBs 9 reinforcement learning algorithms. Train Decision Transformer, Q-Learning, SARSA, and Actor-Critic models. Deploy these plugins to build self-learning agents, implement RL workflows, and optimize agent behavior through experience. Apply offline RL for safe learning from logged data.
