3. Data, Science & AI

Provides standardized prompt engineering and context engineering principles for AI agents, optimized for clarity and signal-to-noise ratio based on Anthropic best practices.

Delegates complex AI/ML tasks to sub-agents for parallel execution, enabling GPU kernel optimization and performance profiling.

Enables users to implement academic research papers from arXiv by providing links or descriptions, focusing on reproducing AI/ML models and algorithms.

Fetches and caches A-share stock data to local CSV, enabling efficient analysis and avoiding context overload for large datasets.

Enables grammar-constrained LLM generation for structured outputs with categorical grammar and type-safe templates.

Provides 16 composable atomic blocks across four layers (Assessment, Transformation, Refinement, Composition) for building customizable AI prompt workflows with progressive disclosure.

Orchestrates stateful multi-agent AI systems with graph-based workflows, conditional routing, cyclic patterns, and human-in-the-loop capabilities.

Provides a DSL for constraint-guided LLM generation, ensuring structured outputs with grammar constraints and logical conditions for reliable AI responses.

Dynamic prompt registry enabling meta-prompts, composition, and quality tracking for AI application development.

Implements Recursive Meta-Prompting (RMP) with categorical syntax for iterative prompt improvement and quality-gated generation in AI systems.

Systematic framework for quantitatively evaluating AI prompt quality against MATH, GSM8K, and Game of 24 benchmarks.

Provides a Python library for implementing compositional distributional semantics and quantum-inspired NLP models using string diagrams and monoidal category theory.

Compositional prompt optimization for LLMs using categorical signatures, module chaining, and automated tuning in declarative AI program development.

Optimizes language model prompts using gradient-based methods and enriched category theory for continuous quality improvement in prompt systems.

Category theory framework for AI meta-prompting, supporting functor, monad, and comonad operations with quality tracking.

Provides polynomial functor implementations for modeling machine learning systems as categorical structures, enabling compositional design of learning components.

Searches and retrieves memories from Nowledge Mem knowledge base with progressive disclosure for past interactions and saved information.

Enforces strict epistemic discipline on Claude to eliminate cognitive biases, ensuring absolute honesty, holistic context review, and structured documentation in all interactions.

Extracts and analyzes large PDFs using semantic chunking and caching for efficient querying of technical documents like NIST and AWS guides.

Extracts and analyzes large Excel workbooks (1MB-50MB+) while preserving formulas, formatting, and complex structures for financial modeling and compliance matrices.

Reduces token consumption 70-80% in Chrome DevTools MCP using AI snapshot strategies and Gemini Flash vision processing for efficient development workflows.

Processes large Word documents (1MB-50MB+) with token optimization, preserving formatting and structure for policy analysis, contract review, and technical reports.

Manages database schema design, data migrations, and ensures data integrity for effective data storage solutions.

Analyzes AI coding session telemetry to identify usage patterns, token efficiency, and workflow optimization opportunities for current or historical data.