Skill UI

TorchDrug is a PyTorch-native toolbox for building graph neural networks tailored to molecules, proteins, and biomedical knowledge graphs, covering property prediction, generation, retrosynthesis, and knowledge graph reasoning with curated datasets and models.

This skill utilizes the Kronos model to perform advanced financial time-series forecasting. It generates quantitative base predictions and then integrates news sentiment analysis to adjust and refine the forecast, making it an essential tool for finance professionals needing actionable market insights.

A comprehensive tool for managing annotation datasets, supporting full CRUD operations, label management, and advanced export functionality. It specifically handles COCO format export and facilitates uploading datasets to platforms like Kaggle, streamlining the data pipeline for AI model development.

A comprehensive tutor designed to build robust statistical thinking. It guides users through the entire data analysis lifecycle, from exploratory data visualization and descriptive statistics to complex hypothesis testing and regression modeling. The focus is on understanding the 'why' and 'how' of statistical logic using real datasets and practical problems, rather than just memorizing formulas. It is tool-agnostic, supporting concepts learned in Python, R, SPSS, or Stata.

This guide details how to set up and execute large-scale, distributed GPU training jobs on the CoreWeave platform. It covers both single-node multi-GPU setups and multi-node training utilizing PyTorch DDP and Kubernetes resources (A100/H100). This workflow is ideal for fine-tuning large language models (LLMs) or running complex deep learning models that require high-performance, shared-storage compute clusters.

This guide demonstrates how to deploy GPU workloads on the CoreWeave platform using Kubernetes (kubectl). It covers deploying vLLM inference services for large language model (LLM) testing and running batch CUDA jobs for benchmarking. This is ideal for users running their first GPU workloads, testing inference pipelines, or verifying cluster access on bare-metal GPU nodes (A100, H100).

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.

This advanced skill audits the integrity of AI/LLM experimental results. It employs a rigorous cross-model review process to analyze evaluation scripts, result files, and published claims. Its purpose is to detect common sources of academic fraud, such as using synthetic ground truth, score normalization manipulation, referencing non-existent results, or misrepresenting the scope of the experiment, thereby ensuring the scientific rigor and trustworthiness of reported findings.

A comprehensive skill designed to query and download large-scale public cancer imaging data from the NCI Imaging Data Commons (IDC). Utilizing the idc-index package, users can access diverse modalities, including radiology (CT, MR, PET) and pathology datasets. The tool supports complex metadata querying, subset selection by cancer type or anatomical site, and visualization, making it essential for AI model training and scientific medical research. No authentication is required for data access.

This guide provides a comprehensive reference architecture for building production-ready, enterprise-grade applications on Palantir Foundry. It covers the entire data lifecycle, detailing best practices for the standard data pipeline (Raw > Clean > Model), Ontology design, external API integration patterns, and robust multi-layered security models. Use it when designing, planning, or optimizing complex data infrastructure within Foundry.

A comprehensive guide to diagnosing and enhancing search relevance within Qdrant. This skill addresses reported issues such as low precision, low recall, or irrelevant matches. It guides users through determining if the problem lies in the embedding model, Qdrant configuration, or the query strategy. Techniques covered include optimizing chunk splitting, tuning HNSW parameters, implementing hybrid search, and utilizing reranking for superior result quality.

Provides a systematic guide for diagnosing and improving poor search quality in vector databases like Qdrant. This guide covers troubleshooting issues such as low recall, when approximate search degrades, selecting optimal embedding models, and correctly tuning HNSW parameters, quantization, and filtering strategies. Essential for optimizing RAG and semantic search pipelines.