Skill UI

nanoGPT is a simplified, minimalist implementation designed specifically for educational purposes. It reproduces the core architecture of GPT-2 (124M) using clean, hackable code, allowing users to understand the entire transformer pipeline from scratch. The framework supports full workflows, including data preparation on various datasets (e.g., Shakespeare, OpenWebText), model training, and text generation, making it ideal for students and researchers learning NLP and deep learning principles.

Automates splitting datasets into training, validation, and test subsets by generating and running Python code so you can quickly prepare data for machine learning model training and evaluation.

Automates the end-to-end training of machine learning models by analyzing datasets, selecting suitable algorithms, tuning parameters, performing cross-validated training, and emitting performance metrics plus a saved model artifact.

Query and download the NCI Imaging Data Commons public cancer imaging datasets via idc-index, filter by metadata, fetch DICOM files, visualize series, and respect licensing details for AI training or research.

This skill provides zero-shot time series forecasting using Google's TimesFM foundation model. It forecasts any univariate time series (like sales, sensor data, or weather) without requiring custom model training. The model delivers point forecasts along with calibrated prediction intervals (quantiles), offering a robust and probabilistic view of future trends. It is ideal for rapid, reliable forecasting across diverse time-based datasets.

bdistill transforms AI interactions into a structured, queryable knowledge base. It extracts, structures, and quality-scores domain knowledge from both closed-source AI models (in-session) and local open-source models (via Ollama). Users can build comprehensive reference datasets, Q&A pairs, or generate structured training data for traditional machine learning, ensuring validated and exportable domain expertise.

This expert tool provides deep, systematic auditing of datasets to assess completeness, consistency, accuracy, and structural validity. It performs full data profiling, detects anomalies (outliers), analyzes missing data mechanisms (MCAR/MAR/MNAR), and calculates a Data Quality Score (DQS). Use it when validating data for model training, debugging pipelines, or establishing ongoing monitoring checks.

Provides programmatic and standardized access to massive, versioned datasets of single-cell and spatial transcriptomics from CZ CELLxGENE. This skill allows users to query millions of cells, analyze gene expression, and access pre-calculated embeddings across various tissues, diseases, and species without downloading entire datasets. Essential for large-scale bioinformatics analysis, machine learning model training, and comprehensive cross-dataset comparison.

Histolab is a specialized Python library for digital pathology designed to process Whole Slide Images (WSI). It automates critical steps including sophisticated tissue detection, customizable tile extraction (random, grid, score-based), and stain normalization. It is essential for preparing high-quality, structured datasets for deep learning model training and comprehensive quantitative whole slide analysis.

This skill guides the secure handling and lifecycle management of large-scale data artifacts, including training datasets, model weights, and checkpoint snapshots, within GPU cloud environments. It provides best practices for data pipelines, covering secure import via Kubernetes PVCs, compliant export, and stringent data validation, ensuring adherence to encryption standards (AES-256) and regulatory requirements (PII masking, RBAC).

A comprehensive tool for creating high-quality training datasets for custom object detection and segmentation models. It supports multiple advanced annotation methods including BBox drawing, SAM2 pixel-perfect segmentation, and DINOv3 visual grounding. Features include AI auto-detection, object tracking, and standardized COCO format export, streamlining the entire data pipeline from capture to training.

Utilizes the idc-index package to query and download large-scale, public cancer imaging datasets from the NCI Imaging Data Commons. This skill allows users to select image subsets based on diverse metadata (e.g., cancer type, modality, anatomical site) and download raw DICOM files for advanced AI training, medical research, and visualization. No authentication is required for data access.