3. Data, Science & AI

Provides implementation patterns for building Gemini AI agents using ReAct loops, Python tool definitions, and LangChain/LangGraph frameworks.

Provides design patterns for coordinating multiple AI agents in complex workflows, supporting architectures like Hierarchical, Peer-to-Peer, and Collaborative.

Provides foundational cognitive architectures for Gemini agents, enabling implementation of reasoning techniques like ReAct and Chain-of-Thought within agent orchestration.

Expertise in debugging MediaPipe pose detection, adjusting confidence thresholds, and resolving landmark tracking issues for accurate video-based pose analysis.

Pulls user trading activity data from Polymarket's API for any specified user.

Framework for defining precise task specifications, deliverables, and negotiation loops for AI agents to reduce ambiguity in complex workflows.

Dispatches subagents in parallel for independent problem domains, optimizing investigation without shared state or sequential dependencies.

Guides video analysis workflows for athletic movements like CMJ and drop jumps, featuring pose detection debugging and kinemotion CLI command execution.

Establishes consistent communication and behavior patterns for the AI agent throughout a session, defining tone, proactivity, and request handling strategies.

Enables seamless querying, filtering, and pagination of Airtable records using the Airtable.js library for data integration.

Validates CMJ and drop jump metrics against physiological norms, including RSI, jump height, and ground contact time.

Creates or enhances AI agents using templates for business roles, consultants, and hub agent configurations.

Enables CLI interaction with Google's NotebookLM for querying and managing knowledge base documents through AI-powered chat.

Seamlessly execute, inspect, and convert Jupyter notebook cells within Claude Code for data science workflows and analysis.

Expertise in debugging, adjusting confidence thresholds, and fixing landmark tracking issues in MediaPipe pose detection pipelines.

Reads and summarizes large documents with context-aware analysis for efficient information extraction.

Enables application integration of AI/LLM capabilities, ML model processing, and generative workflows through API implementations.

Searches past Claude Code conversations for facts, patterns, decisions, and context using semantic or text search.

Automates creation, alteration, and validation of Snowflake semantic views using Snowflake CLI, ensuring DDL correctness and integration.

Converts Product Requirements Documents (PRDs) into the structured prd.json format required by the Ralph autonomous agent system.

Comprehensive tool for creating, editing, and analyzing spreadsheets with formulas, formatting, and data visualization for multiple spreadsheet formats.

Automates extraction of distance data from billing Excel files and updates regional distance cache for Hefei or Jiangxi via automatic region detection.

Generates EVOLEA brand images using reinforcement learning for AI-driven image synthesis.

Executes a wrapper for multi-backend AI code generation, supporting Codex, Claude, and Gemini with file references and structured output.