Data analysis, machine learning, LLM apps, and prompt design (9.1K skills)
Add text and image moderation to Python applications using Azure AI Content Safety. Analyze harmful content by category and severity, manage custom text blocklists, and use the results to apply thresholds for user submissions or AI outputs.
Implement document extraction in Java with Azure AI Document Intelligence. Read text, tables, and fields from PDFs, invoices, receipts, and IDs, or build custom models and classifiers that identify document types.
Create Python agents on Azure AI Foundry that retain conversation context and use multiple tools. The skill covers code execution, file and web search, external tool connections, streaming responses, and outputs organized into defined fields.
Manage Azure AI Foundry project resources from Java. This skill guides authentication and specialized clients for connections, datasets, model deployments, search indexes that organize searchable content, and evaluations used to assess AI models.
Build image analysis features in Java with Azure AI Vision. Add image descriptions, OCR for reading text in images, object and people detection, content tags, and suggested crop regions for thumbnails.
Build anomaly detection features in Java for time-series data, meaning measurements recorded over time. Use Azure AI Anomaly Detector to check individual signals, analyze related signals together, detect trend changes, and manage trained models.
Build text and image moderation features in Java with Azure AI Content Safety. The skill guides category-based harm analysis, severity scoring, and custom blocklists for checking content against terms you specify.
Implement text analysis in Python with Azure AI Language. Covers sentiment detection, extracting names and key phrases, identifying languages, masking personal information, and extracting medical terms, with examples for batch processing and asynchronous clients.
Build document analysis into .NET applications using Azure AI Document Intelligence. Extract text, tables, and typed fields from common documents, then create and manage custom models or classifiers for specialized forms and document categories.
Build .NET voice assistants that exchange audio with Azure AI in real time. This skill guides session setup, speech turn detection, audio and text responses, and function calls that connect conversations to external services.
Build Azure OpenAI features into .NET applications with guidance on authentication, chat, image generation, and audio processing. Examples also cover search integration and structured responses that follow a specified data format.
Turn existing KQL or IRQL query results into relationship diagrams in Kusto Explorer. Generates mappings for entities and connections, adds icons, and composes stored functions to group nodes or add available information to the graph.
Turn tabular Kusto query results into graphs that represent entities and their relationships. Generate KQL, Kusto's query language, to find patterns, shortest paths, and connected groups, with reusable graph models and table exports also covered.
Write, debug, and review KQL queries for log analysis and data exploration. Get guidance on joins, data types, dates, and memory use to address common errors and keep query results manageable.
Detect people, vehicles, animals, and other objects in live camera feeds, with boxes showing their positions, labels, and confidence scores. Adjust target classes, model size, and processing frequency to suit the available hardware.
Detect people, cars, pets, and other objects in live camera frames using a Google Coral Edge TPU through Windows WSL2. Choose from 80 object categories and adjust detection confidence, processing rate, and input resolution.
Evaluate vision-language models, AI that interprets images and text, on smart home video anomaly detection. Sample multiple frames from clips across seven categories and generate HTML reports with accuracy metrics, error breakdowns, and model comparisons.
Fine-tune YOLO26 object detection models with custom labeled image datasets in COCO format. Track training progress, validate detection accuracy, export for the available hardware, and optionally replace the active detection model in Aegis with the trained version.
Select objects in video frames with clicks or boxes using Segment Anything 2. Create masks that mark object areas, propagate selections across a clip, and support labeling and dataset creation within Annotation Studio.
Detect people, cars, and other objects in Aegis camera frames using OpenVINO inside Docker. Adjust detection settings for Intel GPUs, NCS2 hardware, or a CPU, and return object locations and confidence scores.
Create training datasets for custom object detection models by labeling images and video frames. Use AI-suggested boxes, click-based object masks, and tracking, review the annotations manually, and export the dataset in COCO, YOLO, or VOC format.
Detect people, cars, pets, and other objects in live camera frames using a Google Coral Edge TPU on macOS or Linux. Select from 80 object categories and receive results with confidence scores and object locations.
Transform live camera feeds into colorized depth maps that show how near or far objects are. Choose a depth-only view, an overlay, or a side-by-side comparison to inspect spatial layout and activity.
Manage labeled datasets in Aegis Annotation Studio, including creating and deleting datasets, saving and loading annotations, and organizing labels. Export annotations in COCO format, a standard dataset structure, or upload datasets to Kaggle.