Skill UI

NeMo Curator streams GPU-accelerated curation for LLM training, with super-fast fuzzy deduplication, multimodal quality filters, semantic dedupl, PII redaction, NSFW checks, and RAPIDS-scaled pipelines for prepping massive corpora.

Ray Data offers scalable distributed data processing for ML workloads, streaming across CPU/GPU clusters and integrating with PyTorch, TensorFlow, and Ray Train. Use it for batch inference, ETL, and multi-modal preprocessing from a laptop up to hundreds of nodes.

Probabilistic deep-learning framework built on PyTorch for single-cell omics; handles batch correction, integration, multimodal and spatial data, differential expression, and transfer learning with unified APIs.

A comprehensive skill for ingesting, analyzing, and manipulating multi-modal video and audio content. Supports real-time desktop capture, live stream monitoring (RTSP), and file uploads (URL/local). Capabilities include advanced indexing (visual, semantic, temporal), precise moment searching, professional timeline editing (subtitles, overlays, dubbing), transcoding, and generating actionable alerts. Ideal for content creators, monitoring systems, and complex media workflows.

Provides a detailed, multi-platform workflow to monitor long-running computational experiments. It checks the status of jobs across various environments (SSH, vast.ai, Modal), collects screen logs, fetches structured JSON results, and pulls detailed training metrics (loss curves, accuracy) from Weights & Biases (W&B). Essential for MLOps and automated result review.

MarkItDown is a lightweight utility designed to convert heterogeneous documents (PDF, Office files, HTML, etc.) into structured, plain Markdown. It is specifically optimized for text analysis, search indexing, and preparing data for Large Language Models (LLM) and Retrieval-Augmented Generation (RAG) systems, prioritizing content structure over visual fidelity. Supports local, stream, and advanced multimodal extraction scenarios.

A comprehensive research toolkit for building and auditing reproducible workflows for physiological time-series data. It specializes in preprocessing, event/interval analysis, multimodal alignment, and extracting variability and complexity features from signals like ECG and EDA. Designed strictly for academic and research use, emphasizing rigorous data contracts and method-aware validation, while strictly disclaiming any diagnostic or medical device validation role.

scvi-tools is a comprehensive Python framework utilizing deep generative models and variational inference for analyzing diverse single-cell data modalities. It excels in complex analyses such as probabilistic batch correction (scVI), multi-modal integration (MultiVI), and differential expression analysis with uncertainty. It supports single-cell RNA-seq, ATAC-seq, and various multi-omics datasets, providing advanced tools for modern genomic research.

A comprehensive skill for ingesting, analyzing, and manipulating multi-modal video and audio content. Supports real-time desktop capture, live stream monitoring (RTSP), and file uploads (URL/local). Capabilities include advanced indexing (visual, semantic, temporal), precise moment searching, professional timeline editing (subtitles, overlays, dubbing), transcoding, and generating actionable alerts. Ideal for content creators, monitoring systems, and complex media workflows.

A comprehensive strategic guide for authors targeting the ACM International Conference on Multimedia Retrieval (ICMR). This skill helps authors assess the fit of their multimedia retrieval, cross-modal search, or representation learning work for the venue. It provides specific advice on framing the contribution, defining necessary evidence (e.g., ranking metrics, baselines), and structuring the manuscript to meet ICMR's high standards.