Microsoft AI / ML / KM Solution Accelerators
Author: Fatos Ismali
Many Models Solution Accelerator microsoft/solution-accelerator-many-models (github.com)
The Many Models SA is a great starting point for training multiple models in parallel using the ParallelRunStep feature of Azure ML. The accelerator available on the Github link above walks you through a Forecasting scenario using Azure Automated ML. You can also incorporate your own models (e.g. Keras, PyTorch, Croston, AutoARIMA, etc). You’ll leverage Azure Machine Learning predominantly to parallelize training and inferencing.
Knowledge Mining Solution Accelerator microsoft/Accelerator-AzureML_CognitiveSearch (github.com)
The Knowledge Mining SA walks you through the process of creating a knowledge mining solution to enrich your data by identifying custom entities in a corpus of data using an AI custom skill. You’ll leverage a number of capabilities in Azure Cognitive Search and Azure Machine Learning to index unstructured documents, extract entities from documents, apply cognitive skills and many more.
Customer Service Conversational Insights https://github.com/microsoft/Customer-Service-Conversational-Insights
The Customer Service Conversational Insights solution accelerator extends from the previous Knowledge Mining one by focusing on scenarios where you have to deal with chat log, email, or call transcription data. For call transcription data, it uses the generated call transcription schema from the Speech to Text Solution Accelerator below.
Speech to Text Solution Accelerator cognitive-services-speech-sdk/guide.md at master · Azure-Samples/cognitive-services-speech-sdk (github.com)
The Speech to Text solution accelerator once deployed, listens for new .mp3 or .wav files that get uploaded to a designated storage account in your Azure subscription and immediately queues them up for transcription. During deployment of the Solution Accelerator, you have the option to specify whether you want to extract entities, sentiment, mask PII data, and store the resulting information in a SQL database for which the SA provide two PowerBI pre-build dashboards with pre-configured metrics (Speech Insights and Sentiment Insights).
MLOps on Azure microsoft/MLOps: MLOps examples (github.com)
If you’re getting started with MLOps on Azure then the git repo above is a great starting point to kick off your MLOps journey. It contains step by step examples on how to orchestrate a training / inferencing pipeline using Azure DevOps and Azure ML and how to manage an end-to-end MLOps lifecycle.
Video Anomaly Detection Solution Accelerator microsoft/MLOps_VideoAnomalyDetection: Operationalize a video anomaly detection model with Azure ML (github.com)
This is an example of operationalizing a Video Anamoly Detection model using Azure ML.
Deep Learning for Seismic Imaging and Interpretation microsoft/seismic-deeplearning: Deep Learning for Seismic Imaging and Interpretation (github.com)
For geophysicists and data scientists who want to run seismic experiments using state-of-art DSL-based PDE solvers and segmentation algorithms on Azure. The git repo above provides sample notebooks, data loaders for seismic data, utilities, and out-of-the-box ML pipelines.
Recommenders microsoft/recommenders: Best Practices on Recommendation Systems (github.com)
Example Jupyter notebooks on best practices for building recommendation systems. Some algorithms covered are: ALS, SVD, BRP, BiVAE, FastAI and many more.
Here are two examples of building an end-to-end recommendation system for a Retail sector scenario. Both leverage Azure Synapse Analytics and Azure ML.
Commodity Price Prediction Solution Accelerator microsoft/Azure-Synapse-Solution-Accelerator-Commodity-Price-Prediction (github.com)
Forecast prices of commodities and detect anomalies in time-series data using Azure Anomaly Detector and Azure ML.
Customer Revenue Growth Factor Solution Accelerator microsoft/Azure-Synapse-Solution-Accelerator-Financial-Analytics-Customer-Revenue-Growth-Factor: This accelerator was built to provide developers with all of the resources needed to build a solution to identify the top factors for revenue growth from an e-commerce platform using Azure Synapse Analytics and Azure Machine Learning. (github.com)
Identify top factors for revenue growth for an e-commerce platform using Azure Synapse Analytics and Azure ML.
NLP Best Practices and Recipes microsoft/nlp-recipes: Natural Language Processing Best Practices & Examples (github.com)
A comprehensive set of tools and examples in the form of Jupyter Notebooks that cover advanced NLP techniques in areas such as Text Classification, Text Summarization, Question Answering, Entailment, Embeddings, etc.
Verseagility microsoft/verseagility: VERSEAGILITY — NLP Toolkit (github.com)
An example of implementing a NLP solution end to end using a scalable micro-services based architecture and open-source NLP algorithms supporting use cases such as: binary, multi-class and multi-label classification, NER, Question answering, and text summarization. Contains a live demo available here: https://aka.ms/nlp-demo
Computer Vision Recipes microsoft/computervision-recipes: Best Practices, code samples, and documentation for Computer Vision. (github.com)
A set of examples and best practices on leveraging computer vision techniques for scenarios such as: Classification, Object detection, Similarity, Segmentation, Tracking, Crowd counting and many more. All available in Jupyter notebooks.
