3. Data, Science & AI

Queries news articles via Neo4j using fulltext and vector search for semantic content retrieval.

Provides unified file I/O and dataset handling for local, S3, and Hugging Face storage using Unibox utilities, supporting AI/ML data operations.

Orchestrates batch analysis of company 8-K filings, executing prediction and attribution loops with accuracy tracking for earnings insights.

Comprehensive reference for Neo4j schema including labels, relationships, data types, and indexes to explore database structure and financial knowledge graph.

Queries XBRL financial metrics (e.g., EPS, Revenue) from Neo4j for structured data retrieval from 10-K/10-Q filings.

Analyzes stock price movements following 8-K earnings filings using AI to determine primary drivers of market reactions.

Fetches company, sector, industry, and financial metrics data (dividends, splits) from Neo4j graph database for market information.

Enforces evidence citation and domain scope standards for Neo4j subagents to prevent AI hallucinations and maintain audit trails.

Enables multi-step Chain-of-Thought reasoning for answering 'why' questions and comparative analysis with AI-driven explanations.

Analyzes stock price movements post-8-K earnings filings to identify primary drivers using AI-powered attribution.

Compares unmodified responses from Claude, OpenAI, and Gemini by executing identical prompts in parallel.

Predicts stock price direction and magnitude at earnings report release using point-in-time data, prior to earnings attribution analysis.

Prevents 'column does not exist' errors by verifying correct database field names for SQL, Supabase, and Prisma operations.

Analyzes response quality and incorporates user feedback to enhance AI memory system and pattern recognition for continuous improvement.

Enables semantic search and time-based recall of past conversations, decisions, and project contexts across sessions for seamless user interaction.

Answers work-related questions by synthesizing data from 100+ enterprise applications and citing sources.

Provides a collection of skills for context engineering, multi-agent architectures, and production agent systems in AI development.

Summarizes ekkOS's learned patterns and memory operations upon user request, enhancing AI system transparency.

Creates and configures custom sub-agents for Claude Code, managing agent files via .claude/agents directory and command triggers.

Analyzes tabular data (CSV, Excel) to generate visualizations and summary reports for actionable insights.

Generates Korean-language prompts for AI agents, MCP servers, and coding tasks based on user requests.

Routes AI tasks to optimal agents by analyzing task type, scale, and agent capabilities for efficient system operation.

Builds AI agents using Google's Agent Development Kit (ADK) in Python, supporting LLM agents, workflow agents, and multi-agent systems for Vertex AI deployment.

Repomix packages code repositories into AI-optimized files with customizable formats, preserving structure for efficient AI analysis and documentation.