3. Data, Science & AI

Searches Met Museum Open Access Paintings collection using semantic and image-based queries to find similar artworks and match descriptions.

Optimizes documentation, changelogs, and MCP tasks for minimal token usage without quality loss using AI techniques.

Bootstraps AI Developer Workflows (ADWs) infrastructure for programmatic agent orchestration, reusable templates, and observability.

Generates custom Claude Code subagents with YAML configuration, tool restrictions, and system prompts for domain-specific AI assistants.

Starts a new learning episode in a self-learning memory system to track task execution patterns for adaptive AI improvement.

Enables episodic memory retrieval and pattern analysis via the Memory MCP server, including debugging and validation of MCP implementations.

Intelligently plans and coordinates multiple agents for complex multi-step tasks with dynamic optimization and quality gates.

Completes and scores learning episodes to extract patterns and update system heuristics for enhanced future learning and decision-making.

Retrieves past episodic context from memory to inform current decisions, leveraging patterns and solutions from similar tasks for enhanced problem-solving.

Provides domain-specific testing patterns for episodic memory operations in AI systems, covering episode lifecycle, pattern extraction, reward scoring, and retrieval.

Provides medical safety evaluation for AI models using pass^k metrics, failure taxonomy, and anti-gaming graders to ensure robustness.

Runs HealthBench Hard benchmark evaluations for healthcare AI models using multi-specialist council architecture and deterministic safety protocols.

Tests AI model safety persistence in crisis scenarios by evaluating responses to user barriers, minimized distress, and disengagement attempts.

Provides reference for Vercel AI SDK to build AI applications, including text generation, chatbots, and agent integration with major AI providers.

Builds interactive, no-code data dashboards using Lumen YAML, connecting multiple data sources and creating responsive visualizations without coding.

Enables natural language data exploration with RAG and LLM-generated SQL/visualizations for conversational analytics interfaces.

Protocol for recursively improving AI agent methodologies via self-observation and standard refinement.

Reference taxonomy for selecting reasoning modes, inference patterns, and problem-solving methods in AI systems.

Provides defect taxonomy, schemas, and closed-loop remediation for agent-native networks to eliminate avoidable work and risk through evidence-based counterexample analysis.

Centralizes AI model configuration with triggers for model definitions, pricing, capabilities, and routing patterns.

Merges full-text and vector search results using Reciprocal Rank Fusion (RRF) for improved accuracy in Convex-based search applications.

Centralized LLM prompt management with structured storage, extraction, and parameter injection for efficient AI operations.

Configures multi-provider LLM reasoning models with dynamic routing based on triggers and budget constraints.

Provides backend patterns for managing AI generation costs and efficiency, including token tracking, context window management, and rate limiting strategies.