3. Data, Science & AI

Identifies causally responsible model components (layers, heads, positions) via activation patching for transformer model interpretability.

Enables comprehensive spreadsheet creation, editing, and data analysis with formula support and visualization capabilities.

Converts JSON to a compact, human-readable format that minimizes tokens (~40% reduction) and enhances LLM comprehension accuracy.

Compresses verbose LLM prompts by removing redundancy, reducing token count by 40-60% for efficient processing.

Converts verbose text and markdown into token-efficient formats to reduce LLM context size and optimize content for AI processing.

Runs Google's Gemini AI model via CLI for direct queries, Claude vs Gemini comparisons, and task delegation.

Compares 3D CAD models using boolean operations and similarity metrics (IoU, Dice, precision/recall) for model evaluation and ML training.

Applies Edward Tufte's data visualization principles to create clear, impactful presentation slides and charts from data.

Manages decision history with semantic search (Voyage AI) and pattern extraction using ChromaDB for AI-driven learning loops.

Provides PostgreSQL-specific patterns for database design, query optimization, and transaction management.

Optimizes large-scale data processing (50+ items) via sandboxed execution, pagination, and CSV/log analysis to achieve 98%+ token savings.

Applies advanced prompting techniques including chain-of-thought, multishot examples, and Claude 4 patterns to enhance AI model output quality and reliability.

Selects optimal Claude model (Opus 4.5, Sonnet, Haiku) for tasks considering complexity, reasoning depth, and cost/speed trade-offs.

Provides computer vision capabilities for image processing, object detection, segmentation, and visual analysis using AI models.

Builds and manages ETL pipelines for big data processing using Apache Spark and data warehousing solutions.

Provides foundational Python support for data manipulation, algorithm implementation, and data science workflows using Pandas and NumPy.

Builds and deploys deep learning models including CNNs, RNNs, and Transformers for image classification, NLP, and sequence modeling using TensorFlow and PyTorch.

Performs statistical analysis including hypothesis testing, distributions, and inference for A/B testing and experimental validation.

Provides time-series forecasting and analysis using ARIMA, Prophet, and anomaly detection for demand forecasting and pattern identification.

Enables text processing, sentiment analysis, and NLP tasks such as text classification and named entity recognition using LLMs and NLP frameworks.

Enables data visualization, EDA, and dashboard creation using Matplotlib, Seaborn, and Plotly for data-driven insights.

Provides Q-learning, DQN, PPO, and Gym environments for training AI agents in game AI, robotics, and decision systems.

Optimizes machine learning models through quantization, pruning, and hyperparameter tuning to enhance performance and reduce size.

Builds and evaluates machine learning models including classification, regression, and clustering using scikit-learn.