Amazon Technologies, Inc.

Microsoft Technology Licensing, LLC

Microsoft Corporation

Stefano Stefani

Wei Xiao

Timothy Andrew Rath

Swaminathan Sivasubramanian

Jakub Kulesza

David Alan Lutz

James Robert Hamilton

Jai Vasanth

Rande Anthony Blackman

Bjorn Patrick Swift

Raymond S Bradford

Anurag Windlass Gupta

Stuart Henry Seelye Marshall

Chiranjeeb Buragohain

Vladimir Zhukov

Yan Valerie Leshinsky

Kiran-Kumar Muniswamy-Reddy

Poorna Chand Srinivas Perumalla

Akshat Vig

Somasundaram Perianayagam

Kamal Kant Gupta

Yijun Lu

Tengiz Kharatishvili

Miguel Mascarenhas Filipe

Thomas Albert Faulhaber, Jr

Luis Felipe Cabrera

Grant Alexander Macdonald McAlister

Andrea Olgiati

Vikram Sathyanarayana Anbazhagan

James Christopher Sorenson

Alexandre Olegovich Verbitski

Maximiliano Maccanti

Conor J Cunningham

Aleksandras Surna

Madan Mohan Rao Jampani

Sudipta Sengupta

Leo Parker Dirac

Rama Krishna Sandeep Pokkunuri

Vaibhav Jain

Alexander Johannes Smola

Zohar Karnin

Grant A M McAlister

Alexander Richard Keyes

Siddharth Vivek Joshi

Warren Vincent Barkley

Saileshwar Krishnamurthy

Edo Liberty

Jose A Blakeley

Pejus Manoj Das

Amit Gupta

David Craig Yanacek

Tanmoy Dutta

Adam Douglas Morley

Girish Chander

Yongsik Yoon

Raman Mittal

Ganesh Kumar Gella

Xianglong Huang

Eric N Hanson

Oliver Nicholas Seeliger

Ashish Singh

Marc John Brooker

Tate Andrew Certain

Istvan Cseri

Colin Laird Lazier

Matthew Shawn Wilson

Rui Liu

Gopi K Attaluri

Ashok Kumar

Craig Wesley Howard

Yosseff Levanoni

Gaurav D Ghare

Rahul Sharma

Saurabh Gupta

Douglas Brian Terry

Ramachandran Venkatesh

Craig Steven Freedman

Pietro Perona

Hakan Jakobsson

Ajay Kalhan

Beysim Sezgin

Srikumar Rangarajan

Medhavi Dhawan

Rajaprabhu Thiruchi Loganathan

Roman Goldenberg

Mukesh Karki

Yuri L Volobuev

Naveen Prakash

Tushar Jain

Sandeep Bhatia

Akhilesh Mritunjai

Jun Fang

Rahul Sharma Pathak

Anthony A Virtuoso

Peter Alan Carlin

Joseph Patrick Tighe

Sharatkumar Nagesh Kuppahally

Wei Yu

Ming-Chuan Wu

Ranju Das

A system that implements a scaleable data storage service may maintain tables in a data store on behalf of storage service clients. The service may maintain data in partitions stored on respective computing nodes in the system. The service may support multiple throughput models, including a committed throughput model and a best effort throughput model. A service request to create a table may specify that requests directed to the table should be serviced under a committed throughput model and may specify the committed throughput level in terms of logical service request units. The service may reserve low-latency storage and other resources sufficient to meet the specified committed throughput level. A client/user may request a modification to the committed throughput level in anticipation of workload changes, such as an increase or decrease in traffic or data volume. In response, the system may increase or decrease the resources reserved for the table.

System and method for providing a committed throughput level in a data store

A system that implements a scalable data storage service may maintain tables in a non-relational data store on behalf of clients. The system may provide a Web services interface through which service requests are received, and an API usable to request that a table be created, deleted, or described; that an item be stored, retrieved, deleted, or its attributes modified; or that a table be queried (or scanned) with filtered items and/or their attributes returned. An asynchronous workflow may be invoked to create or delete a table. Items stored in tables may be partitioned and indexed using a simple or composite primary key. The system may not impose pre-defined limits on table size, and may employ a flexible schema. The service may provide a best-effort or committed throughput model. The system may automatically scale and/or re-partition tables in response to detecting workload changes, node failures, or other conditions or anomalies.

System and method for implementing a scalable data storage service

Remote grouping operations may be used to perform queries for distinct values. A query may be received at a query engine that includes a request for distinct items from a data set stored in a data store. To perform the query, a request may be made to the data store to group items from the data set by value when returned from the data store to the query engine. The query engine may then evaluate the items grouped by value that are returned from the data store to determine distinct items as part of performing the query.

Remote grouping operations to perform queries for distinct values

A distributed data store may implement passive distribution encryption keys to enable access to encrypted data stored in the distributed data store. Keys to encrypt a data volume stored in the distributed data store may be encrypted according to a distribution key and provided to a client of the distributed data store. Storage nodes that maintain portions of the data volume may receive the encrypted key from a client to enable access to the data volume. The storage nodes may decrypt the key according to the distribution key and enable access to the data volume at the storage nodes. In some embodiments, a key hierarchy may be implemented to encrypt the keys that provide access to the encrypted data. The key hierarchy may include a user key.

Passive distribution of encryption keys for distributed data stores

Techniques for in-flight scaling of machine learning training jobs are described. A request to execute a machine learning (ML) training job is received within a provider network, and the ML training job is executed using a first one or more compute instances. Upon a determination that a performance characteristic of the ML training job satisfies a scaling condition, a second one or more compute instances are added to the ML training job while the first one or more compute instances continue to execute portions of the ML training job.

In-flight scaling of machine learning training jobs

A determination is made as to whether a value of a first parameter of a first application is to be obtained using a natural language interaction. Based on received input, a first service of a plurality of services is identified. The first service is to be used to perform a first task associated with the first parameter. Portions of the first application to determine the value of the first parameter and to invoke the first service are generated.

Back-end task fulfillment for dialog-driven applications

A non-relational database may be emulated using a relational database with a distributed data store. A request to access the non-relational database may be received at a frontend for a relational database engine that emulates an interface for a non-relational database engine. The request may be translated into the format for the relational database engine and performed by the relational database engine. The relational database engine may access data for the non-relational database at storage nodes that store the data in a relational table column of non-relational data type.

Emulating a non-relational database on a relational database with a distributed data store

Techniques for automated machine learning (ML) pipeline exploration and deployment are described. An automated ML pipeline generation system allows users to easily construct optimized ML pipelines by providing a dataset, identifying a target column in the dataset, and providing an exploration budget. Multiple candidate ML pipelines can be identified and evaluated through an exploration process, and a best ML pipeline can be provided to the requesting user or deployed for production inference. Users can configure, monitor, and adapt the exploration at multiple points in time throughout.

Automated machine learning pipeline exploration and deployment

At a first resource to be used to perform a computing operation, a pair of execution environments is configured. I/O permissions of programs running in the different environments are based on respective sets of constraints. A program performs the operation in one of the environments, with input data being provided to the program from the second environment. A result of the operation is provided to a destination from the second environment.

Isolation techniques at execution platforms used for sensitive data analysis

Techniques for using scoring algorithms utilizing containers for flexible machine learning inference are described. In some embodiments, a request to host a machine learning (ML) model within a service provider network on behalf of a user is received, the request identifying an endpoint to perform scoring using the ML model. An endpoint is initialized as a container running on a virtual machine based on a container image and used to score data and return a result of said scoring to a user device.

Packaging and deploying algorithms for flexible machine learning

Methods, systems, and computer-readable media for debugging and profiling of machine learning model training are disclosed. A machine learning analysis system receives data associated with training of a machine learning model. The data was collected by a machine learning training cluster. The machine learning analysis system performs analysis of the data associated with the training of the machine learning model. The machine learning analysis system detects one or more conditions associated with the training of the machine learning model based at least in part on the analysis. The machine learning analysis system generates one or more alarms describing the one or more conditions associated with the training of the machine learning model.

Debugging and profiling of machine learning model training

Methods and apparatus for providing persistent execution environments for computation systems including but not limited to interactive computation systems. A service is provided that extends the notion of static containers to dynamically changing execution environments into which users can install code, add files, etc. The execution environments are monitored, and changes to an execution environment are automatically persisted to environment versions(s) so that code run in the execution environment can be run later or elsewhere simply by referring to the environment. There is no explicit build step for the user. Instead, incremental changes are added to environment versions which are stored and are ready to be used to instantiate respective execution environments on other compute instances.

Persistent execution environment

A network-accessible machine learning service is provided herein. For example, the network-accessible machine learning service provider can operate one or more physical computing devices accessible to user devices via a network. These physical computing device(s) can host virtual machine instances that are configured to train machine learning models using training data referenced by a user device. These physical computing device(s) can further host virtual machine instances that are configured to execute trained machine learning models in response to user-provided inputs, generating outputs that are stored and/or transmitted to user devices via the network.

Network-accessible machine learning model training and hosting system

Working set ratio estimations of data items in a sliding time window are determined to dynamically allocate storage for the data items. A working set ratio may be determined by accessing a fixed-size array that stores respective timestamps of last accesses of data items to determine which data items are useful to determine an estimate of a working set for the application within a range of time. The working set ratio is then determined from an estimated working set and an amount of computing resources allocated to the application by the estimated working set. The amount of the computing resources allocated to the application may then be automatically scaled according to the determine working set ratio.

Working set ratio estimations of data items in a sliding time window for dynamically allocating computing resources for the data items

A request is received associated with a review. Within first content, a first field of interest and a second field of interest are identified and within second content, a third field of interest and a fourth field of interest are identified. A review is generated that includes a first indication of the first field of interest and a second indication of the second field of interest within the first content, as well as a third indication of the third field of interest and a fourth indication of the fourth field of interest within the second content. The review is transmitted to a device of a reviewer for reviewing the content.

Providing content reviews based on AI/ML output

Growing community of inventors

Issaquah, WA, United States of America

Stefano Stefani