3. Data, Science & AI

Provides expert guidance for data analysis, machine learning model development, and AI integration using Python libraries and data pipelines.

Configures event tracking and analytics implementations using GA4, PostHog, and Google Tag Manager for data collection and measurement.

Analyzes user prompts and classifies them into topic types (instruction, question, context) using natural language processing.

Orchestrates concurrent execution of multiple Claude agents to independently resolve failures or tasks without shared dependencies.

Formats structured data (tables, logs, API responses) into TOON v2.0 for token-efficient use in RAG pipelines and data processing tasks.

Selects the optimal DQX specialized AI agent to efficiently address specific user tasks.

Designs AI agent systems with agent loops, tool calling, planning, memory, and multi-agent coordination for autonomous workflows and task automation.

Builds RAG systems using chunking, embedding models, and vector databases to enable knowledge base integration and semantic search.

Diagnoses ClickHouse INSERT performance issues, optimizing batch sizing and part creation for efficient data ingestion pipelines.

Summarizes large articles from URLs or files via an AI-powered chunking and reduction pipeline, saving concise outputs to markdown files.

Manages persistent context across AI sessions with memory architecture, privacy controls, and efficient retrieval, integrating with the Claude Mem plugin.

Provides incident response and diagnostics for ClickHouse databases, identifying performance bottlenecks and generating root cause analysis reports with actionable insights.

Analyzes ClickHouse query performance, identifies slow queries, and provides optimization recommendations to resolve latency and timeout issues.

Analyzes ClickHouse external dictionaries for configuration, memory usage, reload status, and performance to resolve issues and load failures.

Specializes in Supabase database management including Row Level Security (RLS) policies, schema migrations, and type generation for application integration.

Analyzes ClickHouse table structures, partitioning, ORDER BY keys, and materialized views to identify anti-patterns and optimize schema design for data processing efficiency.

Transforms vague requirements into high-precision AI prompts through guided interaction, supporting system prompt creation, optimization, and command-based generation.

Extracts HEAL matrix content from PowerPoint slides and saves as formatted text files matching N2 HEAL specification.

Transcribes WhatsApp voice notes to text with domain-specific corrections for mining site names.

Extracts key sections from Epiroc BRMO BEV weekly PDF reports into structured markdown format for Week N.

Enables secure querying of production PostgreSQL databases using Entra ID authentication for user investigation, duplicate debugging, and ad-hoc data analysis.

Automates transcription of WhatsApp voice notes with AI-powered correction of common errors in mining site communications.

Executes AI model tasks (Claude, Gemini, etc.) with memory and resume capabilities via stdin protocol for stateful interactions.

Step-by-step workflow for deriving mathematical models and combining formulas, supporting AI/ML model construction tasks.