3. Data, Science & AI

Queries Parquet files with DuckDB, inspects metadata, joins tables, and converts results to pandas for data analysis, preparing datasets for EDA.

End-to-end AutoGluon TabularPredictor workflow for training, prediction, threshold calibration, and sklearn integration.

Builds training sets with reject inference and fuzzy augmentation for tabular data, incorporating PD-based weights to enhance Autogluon model training.

Generates and compares EDA reports with sampling, consistent seeds, and HTML outputs, integrated with Parquet data workflows.

Command-line tool for rendering Quarto Markdown (.qmd) documents into various formats for scientific publishing.

Enables creation, editing, and analysis of spreadsheets including formulas, formatting, and data visualization for data processing tasks.

Builds scikit-learn compatible custom estimators following official guidelines for init, fit, predict, and validation, enabling seamless integration with sklearn wrappers.

Creates interactive Desmos graphs in Obsidian for visualizing mathematical functions, curves, and relationships with customizable styling.

Consults GPT-5.2 via Codex CLI to provide second opinions or alternative perspectives on technical problems.

Fetches YouTube video metadata including title, duration, and description using yt-dlp for data retrieval.

Provides HTTP API operations for Elasticsearch including search, indexing, bulk, scroll, and alias management.

Provides HTTP API for speech-to-text conversion using local whisper.cpp server with CUDA acceleration for efficient audio transcription.

C# utility for creating, updating, and multi-threaded generation of Parquet data files, optimized for efficient data storage and processing.

Executes PostgreSQL DDL, manages schemas, and performs query operations using the PgQuery tool for efficient database management.

Provides RDBMS access patterns for DuckDB, MySQL, PostgreSQL, SQL Server, and DBISAM using ODBC and native drivers.

Enables local inference of Qwen 2.5 LLM via Ollama for text analysis and AI-driven automation tasks.

Fetches GitHub issue details with AI-generated analysis comments, extracting structured data for architecture planning in WescoBar workflows.

Extracts architectural alignment, feasibility, implementation, and testing insights from GitHub AI analysis comments.

Creates AI agent projects with Mastra framework, guiding integration of agents and workflows into TypeScript/JavaScript applications.

Extracts version or uuid fields from YAML frontmatter in Markdown files using the uv tool.

Fetches GitHub code repositories and formats them into LLM-friendly context for analysis and understanding.

Provides structured AI consultation workflows and prompt templates for code review, security audits, and architecture analysis using multiple AI models.

Expertise in migrating content and data between Drupal versions, including D7 to D10, CSV, and JSON API imports.

Provides an AI-driven self-evolution layer that captures usage patterns, learns from errors, and generates reusable knowledge to enhance agent skills over time.