3. Data, Science & AI

Builds AI agents for financial analysis, risk assessment, and investment recommendations using data processing and predictive models.

Orchestrates tasks across multiple AI providers (Claude, OpenAI, Gemini) by delegating to specialized models, routing based on capabilities, and implementing fallback strategies.

Extracts IDs, titles, and traceability metadata (repo, path, hash) from specifications for requirements management and code traceability.

Provides a token-efficient data encoding format (TOON) that reduces token usage by 30-60% compared to JSON for structured data in LLM applications.

Orchestrates tasks across multiple AI providers and execution environments, enabling complex AI workflows via delegation and native invocation.

Generates reusable semantic patches for AI agent prompts by comparing changes, documenting modifications, and sharing best practices across agent skills.

Converts Product Requirement Documents (PRDs) into Ralph's standardized JSON format for seamless agent system integration.

Provides unified package management for scientific Python environments, combining conda and PyPI for speed and reproducibility.

Applies semantic patches to AI agent prompts, enabling updates, migrations, and consistency across agent skills using .agent.patch files.

Creates charts, graphs, and visualizations from data for clear analysis and reporting.

Converts Product Requirements Documents into structured XML specifications optimized for AI coding agents' consumption.

Verifies accuracy and completeness of FireCrawl-scraped Markdown files against original web pages for data integrity.

Guides users in creating specialized Claude Code agents through agent definition files, frontmatter, prompts, and tool scoping for task-specific automation.

Expert assistance for Microsoft SQL Server development, including T-SQL queries, performance optimization, schema design, and Node.js integration via mssql package.

Creates isolated Ralph agents to execute parallel subtasks via Claude's parallel processing capabilities.

Guides installation and management of AI coding tool artifacts including skills, commands, and hooks using AGPM for agent development.

Automates Recce guidance for data validation in dbt projects, triggered by project context or user queries.

Automates syncing of Garmin fitness data including activity, sleep, and heart rate, providing AI-driven health insights.

Optimizes JAX code performance through profiling, hypothesis testing, and benchmarking to improve speed and operations per second.

Builds Milvus vector database collections via natural language prompts, optimized for RAG systems and semantic search applications.

Automates prompt assembly from templates for dispatching coding and verification tasks to AI agents in development workflows.

Automatically scrapes posts from gpters.org member profiles and saves them in Markdown format using Selenium for dynamic page extraction.

Builds self-hosted conversational AI assistants with CopilotKit and LangGraph, enabling agentic backends, streaming, and shared state without hosted services.

AI-powered development assistant using RAG, local LLMs, and DAG task decomposition to autonomously update project artifacts like documentation and code.