3. Data, Science & AI

Applies DE6 taxonomy to categorize entities into hierarchical structures for organized data management and retrieval.

Provides offline text-to-speech synthesis using Sherpa-ONNX, enabling local speech generation without internet or cloud dependencies.

Manages persistent agent memory via beads methodology, orchestrates parallel subagents, and verifies bead accuracy.

Enables building autonomous AI agents through prompt-defined features, eliminating code for agent behavior and supporting self-modifying systems.

Provides a guide for building custom AI agents with proper configuration, tools, permissions, and reasoning for specific tasks.

Manages background agent execution, parallel task handling, and agent orchestration patterns including run_in_background and progress tracking.

Debugs token-level penalty alignment issues in tokenizers, ensuring banned tokens are correctly penalized and resolving edge cases in tokenization.

Enables token-level reward implementation and debugging for TRL's GRPOTrainer, facilitating reward computation adjustments in reinforcement learning for language models.

Creates a new agent for a Multi-Agent System (MAS) with a standardized code structure.

Runs automated evaluation tests to assess the performance and behavior of AI agents in various scenarios.

Generates charts, maps, and hierarchical visualizations in Flutter via the data_visualization package suite.

Creates reusable scripts to fetch, enrich, and process data from Hugging Face API for AI-driven data automation and analysis.

Provides Supabase-recommended best practices for optimizing PostgreSQL database performance, query design, and schema configuration.

Enables A/B testing and performance measurement of AI prompts to optimize output quality and model effectiveness.

Enables querying historical feedback data by topic, source, or theme for analysis and reporting.

Refines and compresses AI instructions for LLM agents to ensure clarity, conciseness, and effectiveness in agent communication.

Orchestrates multi-agent systems by decomposing complex tasks and coordinating specialized AI agents for efficient execution.

Provides iterative modular development framework optimized for token efficiency in LLM-based applications.

Analyzes drilling performance, efficiency, and cost metrics using API 12 well numbering across Gulf of Mexico fields for benchmarking and comparison.

Generates interactive metocean data visualizations including time series, wave roses, and geographic maps for exploration and monitoring.

Statistical analysis of metocean data including extreme value analysis and return periods for engineering design criteria and operational limits.

Analyzes HSE incident data for risk assessment, safety scoring, trend analysis, and ESG-compliant economic evaluation.

Automates startup, verification, and checklist execution for Anviloop's WSL-based ML sidecar summarizer in nightly headless data processing runs.

Fetches real-time and historical metocean data from NDBC, CO-OPS, Open-Meteo, and ERDDAP for marine forecasting and analysis.