3. Data, Science & AI

Coordinates distributed neural network operations using sheaf theory and graph Laplacians for efficient multi-node collaboration in graph-based AI systems.

Guides efficient session data access in ultrawork, optimizing for token efficiency with JSON/Markdown patterns and common best practices for AI agents.

Enables semantic search over vector databases to retrieve relevant documents based on user queries and content matching.

Suggests MOC updates and new MOCs using semantic clustering to organize personal knowledge bases and note clusters.

Provides guidance for creating, modifying, and troubleshooting continuous queries in Drasi using the Cypher query language.

Analyzes knowledge graphs to uncover connections, hubs, orphans, and generate insights for knowledge mapping and graph reports.

Guides database design patterns, SQL best practices, ORM usage, and migration strategies for efficient data operations.

Builds production-ready Tavily API integrations for web search, content extraction, and research in RAG systems and agentic workflows.

Anonymizes CSV files by removing personal identifying information (PII) and adding datetime stamps for data privacy compliance.

Optimizes command prompts and workflow templates using reinforcement learning to enhance effectiveness and user experience in AI-driven systems.

Creates and manages evaluation suites for AI agent performance testing, including coding, conversational, and research agents.

Verifies mathematical content correctness through BDD scenarios, ensuring accuracy in scientific and educational contexts.

Discovers DuckDB data files and verifies absolute paths for reliable data access and processing.

Enables querying past sessions and self-discovery via DuckDB, facilitating cross-session awareness and data-driven introspection.

Creates domain-specific AI skills by processing tool documentation, enabling rapid development of specialized AI capabilities.

Generates linear dynamics models for nonlinear systems using Koopman operator theory from observable data.

Configures LLM gateway routing with model fallbacks, cost/latency-based routing, and A/B testing using OpenRouter and LiteLLM.

Generates and edits images via Gemini API, supporting text-to-image, style transfers, and multi-reference composition for creative projects.

Specialized tool for processing massive context windows (1M-2M tokens) using Gemini CLI with intelligent safety evaluation and usage direction.

Coordinates multi-agent AI workflows (including Gemini and local models) via capability-based delegation and Shell-As-Bus pattern.

Creates goal-oriented LLM prompts and agent instructions using role, objective, and latitude patterns for effective AI communication.

Query and analyze LLM traces, prompts, and metrics for debugging, performance optimization, and model comparison.

Manages all LLM development workflows including model configuration, prompt engineering, and API integration.

Generates visual diagrams for OSCAL documents, including control hierarchies and component relationships, to enhance understanding of security and privacy controls.