데이터·AI 스킬

Combine capabilities from multiple fine-tuned LLMs without retraining using mergekit. Choose a merging method, configure how models are blended, and evaluate the resulting model on tasks such as math, code generation, and conversation.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Combining math, coding, and conversation capabilities
• Blending two models at different ratios
• Comparing variants made with different merging methods

Prepare training datasets for large language models with GPU-assisted cleaning. Covers duplicate removal, quality filtering, and personal information redaction, with workflows for text, images, video, and audio using NeMo Curator.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Clean web data for model training
• Remove exact and near duplicates from large text collections
• Redact personal information from training text

Get guidance on distributed training with PyTorch FSDP, which splits model parameters across devices. The skill covers implementation, mixed precision, moving data to CPU memory, communication setup, and debugging distributed training code.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Implement distributed training with FSDP
• Review mixed precision and CPU offloading settings
• Choose a distributed communication backend

Design autonomous AI agents that break goals into tasks, use tools, and review their own work. Covers reasoning and action loops, separate planning and execution, and reliability measures such as cost limits, validation, and logging.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Choosing how an agent plans and executes tasks
• Breaking goals into constrained steps
• Adding self-review and correction to an agent

Build and train Mixture of Experts models, which activate selected specialist networks for each input. Use DeepSpeed or HuggingFace workflows to configure routing, balance workloads, distribute experts across GPUs, and optimize inference.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Designing MoE models for a limited compute budget
• Implementing Mixtral-style expert routing
• Adjusting expert workloads and training stability

Design memory systems that help AI agents retrieve relevant past information. Choose short-term and long-term memory structures, compare vector storage options, and review document splitting and retrieval strategies for different kinds of knowledge and experience.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Design memory for information used across conversations
• Choose storage for experiences, facts, and procedures
• Review retrieval when an agent misses saved information

Extend the amount of text a language model can process at once. Compares RoPE, YaRN, ALiBi, and position interpolation, with implementation examples, additional training guidance, and memory optimization approaches for long document processing.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Extend an existing model to handle longer inputs
• Compare methods for representing token positions
• Plan additional training with long documents

Design collaborative AI teams with CrewAI by defining each agent’s role, goals, tasks, and expected results. Covers passing work between agents, manager-led coordination, memory settings, and execution planning for workflows such as research and writing.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Build a team that researches and writes articles
• Coordinate specialists through a manager agent
• Define agent roles and expected task outputs

Manage the limited information an LLM can process at once. Use importance-based summaries, selective trimming, token counting, and content placement to plan what stays in context and address information loss during long conversations.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Planning how to handle conversations near the token limit
• Summarizing a long dialogue while preserving key information
• Reviewing the placement of important instructions and facts

Design and implement systems that collect, transform, validate, and store data for analysis. Covers scheduled and real-time processing, data warehouses, quality checks, monitoring, and cost tuning using tools such as Apache Spark, dbt, and Airflow.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Building scheduled data collection and transformation workflows
• Designing a data warehouse for analysis
• Synchronizing database changes in real time

Analyze business data, test hypotheses, and build predictive models with statistical and machine learning methods. Get support across data exploration, experiment design, visualization, result validation, and planning model deployment and monitoring.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Analyzing customer churn and predicting customers at risk
• Designing an A/B test and evaluating its results
• Forecasting demand for inventory planning

Evaluate code generation models with HumanEval, MBPP, and other benchmarks. Test generated programs, compare models across programming languages, and record pass@k scores, which estimate the chance that at least one of k solutions passes the tests.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Comparing code generation models on shared benchmarks
• Measuring correctness across programming languages
• Evaluating models trained to follow instructions

Configure large language model training with NVIDIA Megatron-Core across multiple GPUs. This skill helps choose how to split computation, set training parameters, monitor GPU efficiency, and address memory limits or unstable training.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Planning multi-GPU training for a LLaMA-style model
• Distributing MoE model experts across GPUs
• Tuning training throughput and GPU utilization

Adapt machine learning training to run across multiple GPUs or machines with Ray Train. Covers changes to existing code, comparing training settings with Ray Tune, and saving checkpoints so training can resume after failures.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Adapting existing PyTorch code for multiple GPUs
• Training HuggingFace models across multiple machines
• Comparing learning rates and batch sizes with Ray Tune

Reduce LLM weight usage through pruning, which removes selected model parameters. Explore Wanda and SparseGPT for compression without retraining, choose hardware-friendly patterns for faster inference, and compare the compressed model's accuracy with the original.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Compressing an LLM without retraining
• Comparing pruning methods
• Exploring faster inference on supported hardware

Build data processing pipelines for machine learning, from preprocessing to batch predictions. Uses Ray Data to handle datasets larger than memory, distribute work across CPUs and GPUs, and feed data into training frameworks.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Preprocessing datasets larger than memory
• Running batch predictions across multiple GPUs
• Preparing image and tabular data for training

Organize PyTorch training code with LightningModule and Trainer. Follow examples for validation, logging, model saving, and distributed training across multiple GPUs, with callbacks that add monitoring or stop training when progress stalls.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Converting an existing PyTorch training loop
• Adding validation, logs, and model checkpoints
• Configuring training across multiple GPUs

Design AI agents that plan tasks, call tools, and retain useful information. Covers coordination between agents, evaluation, and debugging, with practical patterns for execution limits, error handling, and tracing what an agent does.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Designing an agent that plans and executes tasks
• Organizing tool definitions and selection
• Deciding what an agent should remember

Build collaborating AI teams with CrewAI by defining roles, tasks, and shared context. Configure ordered execution or manager-led delegation, add memory and tools, and create workflows that branch based on events or results.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Build a research-to-writing agent workflow
• Coordinate specialists through a manager agent
• Create workflows with conditional branches

Connect a company SQL database, support ticket exports, and documents in cognee, linking shared people, teams, and projects so questions about ownership, decisions, and ongoing work can draw on several sources together.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Finding owners across company databases, tickets, and documents
• Tracing decisions across multiple company sources
• Checking who is handling a support issue

Connect cognee to additional databases and data sources through community packages. The skill covers installing and registering adapters, importing data through connectors, configuring monitoring, and developing extension packages with examples and tests.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Connect cognee to Qdrant or another database
• Import data from Slack or Gmail
• Configure monitoring with Keywords AI

Store conversations, agent activity, and feedback in cognee sessions, then carry them into long-term memory. This skill explains improve(), its processing stages, result statuses, and settings for diagnosing skipped work or controlling automatic runs.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Saving conversations and agent activity by session
• Moving session content into long-term memory
• Investigating skipped or blocked processing stages

Query documents and conversations stored in cognee to retrieve answers, supporting passages, or context for an AI model. This skill guides search method selection, source filtering, result interpretation, and troubleshooting empty or unexpected results.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Searching selected datasets or conversation sessions
• Retrieving supporting passages or model context
• Choosing a search method for a question

Find and remove stored content from cognee using its SDK, HTTP API, or CLI. Inspect documents before deletion, choose document or dataset removal, or keep raw files while clearing derived graph relationships and search vectors.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Find and delete a specific stored document
• Remove an entire dataset
• Clear derived memory while keeping original files