Data Science & ML Agent Skills

LLM Evaluation & Benchmarking

Implements comprehensive evaluation frameworks for LLM applications using automated metrics, human feedback, and LLM-as-judge patterns.

Notebook Checkpoint Pattern

Implements a structured checkpointing system for connecting multiple Jupyter notebooks into robust, memory-efficient data pipelines.

Trading Dual-Folder Workflow

Manages separate development and production environments for private repository trading models and Colab training.

Trading Gates & Pattern Filter

Implements a multi-stage validation system for algorithmic trading that filters signals through pattern-matching and risk-management gates.

Persistent Market Data Cache & Gap-Filling

Optimizes historical market data retrieval by eliminating redundant downloads through persistent caching and incremental gap-filling.

ReasoningBank Intelligence

Empowers AI agents with adaptive learning and pattern recognition to optimize workflows and decision-making strategies through experience.

GPU Correlation Accelerator & Cache

Accelerates financial correlation matrix computations using GPU-vectorized PyTorch operations and persistent SQLite caching.

Jupyter Autoreload Awareness

Optimizes Jupyter notebook workflows by preventing unnecessary kernel restarts when using the IPython autoreload extension.

Notebook-to-Module Refactoring

Refactors Jupyter notebook code into reusable Python modules while preserving critical variable definitions and configuration.

Trading Position Sizing & Action Space

Implements a 7-action space with integrated position sizing and small account simulation for reinforcement learning trading models.

Deconvolution Intensity Scaling

Ensures quantitative accuracy in microscopy deconvolution by preserving original intensity relationships across image channels.

Trading Model Training Gating

Automates the archival and quality classification of algorithmic trading models based on performance metrics and risk thresholds.

Colab Unzip Path Manager

Corrects file path errors and API key configurations in Google Colab environments after repository extraction.

Network Architecture Sizing for Trading RL

Optimizes PPO neural network dimensions to balance trading model capacity, inference speed, and hardware memory usage.

Training Resilience for ML Trading

Optimizes Reinforcement Learning training by preventing premature stops and implementing adaptive recovery for trading models.

Advanced Excel & Spreadsheet Manager

Automates the creation, editing, and analysis of professional Excel spreadsheets with advanced formula support and industry-standard financial modeling.

NCI Manipulation Analysis

Detects narrative manipulation and propaganda patterns in text and URLs using the Narrative Credibility Index (NCI) Protocol.

Weaviate RAG Integration

Implements and manages Retrieval-Augmented Generation (RAG) systems using Weaviate vector databases for semantic search and document retrieval.

Scientific Analysis & Reporting

Generates structured scientific research reports by analyzing raw data, verifying physical models, and calculating quantitative performance metrics.

LLM Application Evaluation

Implement comprehensive evaluation frameworks for LLM applications using automated metrics, human feedback, and benchmarking.

Local LLM & Home Assistant Integration

Integrates local Large Language Models with Home Assistant and Ollama for private, high-performance smart home automation.

Japan Debt & Fiscal Risk Analyzer

Quantifies Japan's interest-to-tax burden through real-time yield analysis and stress testing to identify fiscal risk zones.

Lithium Supply-Demand Radar

Analyzes the lithium supply chain and price regimes to provide data-driven insights for commodity investments and ETF exposure.

Multi-Agent Trading Integration

Integrates multi-agent Claude systems into algorithmic trading pipelines to optimize model training and manage live risk with automated oversight.

Selection Data Caching for Trading

Optimizes financial data retrieval by caching market symbol data to reduce API latency and avoid rate limits during trading bot development.

Sparse Expression Visualization

Optimizes the visualization of sparse single-cell gene expression data by implementing alternative plotting patterns that prevent boxplot collapse.

Trading Drawdown Guardrails

Implements standardized risk management and drawdown protection patterns for algorithmic trading systems.

Vector Awkward HEP Analysis

Simplifies High Energy Physics analysis by integrating Python vector libraries with Awkward Arrays for efficient kinematic calculations.

Backtest Result Persistence

Persists trading backtest results to an SQLite database to enable historical performance comparison and model optimization.

Region-Aware Spatial Cell Matching

Optimizes CODEX and scRNAseq data integration by incorporating tissue heterogeneity into spatial cell matching workflows.