3. Data, Science & AI

Integrates multiple LLMs (OpenAI, Gemini, Ollama) for collaborative multi-agent workflows, supporting role division, consensus, and dynamic scenario configuration to enhance complex task resolution.

Guides integration of local LLM backends (Ollama/GPT-OSS) in AGILAB with correctness-focused prompt engineering.

Synchronizes machine learning model files between servers for consistent deployment and version control.

Analyzes, summarizes, and converts PDF, DOCX, XLSX, and PPTX documents with data extraction capabilities.

Provides GPU-accelerated image loading, memory-mapped dataset handling, and Jupyter notebook visualization for computer vision tasks.

Transforms simple prompts into context-aware, comprehensive prompts for AI interactions. Includes Context7 prompt engineering guides for strategy implementation.

Offers memoized OpenAI client utilities and chat format transformations for efficient LLM integration within speedy_utils.

Deprecated utility for running multiple AI agent threads in parallel with bounded concurrency. Use evolving-workflow instead.

Calibrates survey weights to match population targets, enhancing microdata representativeness for policy analysis in PolicyEngine US.

Enables summing variables across entities in PolicyEngine using the adds attribute and add() function for policy data aggregation.

Provides advanced simulation mechanics for policy analysis using policyengine.py, including data mapping and output handling via ensure(), output_dataset.data, and map_to_entity().

Analyzes policy impacts for U.S. congressional districts using HuggingFace datasets. Handles district codes (e.g., NY-17) and representative queries.

Applies L0 regularization for neural network sparsification and intelligent sampling to optimize survey calibration processes.

Core simulation engine for policy analysis, enabling data-driven calculation of policy impacts through modeling.

Executes PolicyEngine microsimulation for population-level policy impact analysis, calculating winners/losers and weighted national averages.

Enables weighted data analysis for survey microdata, calculating inequality, poverty, and distributional metrics using pandas.

Enables UK tax and benefit policy microsimulation through situation creation and standardized workflow patterns.

Optimizes PolicyEngine calculations using NumPy vectorization patterns, avoiding scalar loops with array operations.

Enhances UK survey data using FRS and WAS imputation patterns to handle missing values and improve dataset integrity.

Enforces standardized variable patterns in PolicyEngine for no hard-coding, federal/state separation, and metadata compliance.

Provides testing patterns for PolicyEngine data generation pipelines, specifically for US and UK data modules.

ML-powered imputation tool for filling missing values in survey datasets, utilized in policyengine-us-data.

Defines structured parameter patterns for PolicyEngine including YAML format, naming conventions, and federal/state metadata organization.

Offers practical tips and best practices for efficient SQLite database management and query optimization.