Microsoft Technology Licensing, LLC

The Brigham and Women's Hospital, Inc.

Google Inc.

USA As Represented by Secretary of the Navy

Harvard College

Yu Shrike Zhang

Ali Khademhosseini

Yun-Cheng Ju

William Chan

Navdeep Jaitly

Ashwarya Poddar

Darren Shaun Firth

Nanda Kishore Sreerama

Lei Lu

Aaron Brady O'Callaghan

James Thomas Godish

Zachary Thomas Capehart

Ying Xu

Dong Yu

Quoc V Le

Chee Heng Chew

James G Droppo

Bryan T Starbuck

Zeke Koch

Michael Lewis Seltzer

Pankaj Sharma

Ray Sun

Rajendra H Vishnumurty

Steven Andrew Cover

Manish Kumar Shukla

Chad M Whitney

Pavel Glozman

James T Jacoby

Xiaohui Wang

Michael Eatherly

Grigoriy A Orlov

Swati Gupta

Chunyan Zhao

Described are systems and methods for in vivo multi-material bioprinting. The in vivo multi-material bioprinting can be used to fabricate biomedical constructs within a patient minimally invasively. The systems and methods can utilize a multi-material bioprinter, which includes a biocompatible portion. The biocompatible portion can include a single printhead for in vivo bioprinting. The single printhead can include a plurality of outlets, each linked to one of a plurality of reservoirs. Each of the plurality of reservoirs can each house a different bioink for bioprinting. Each of the plurality of outlets can be activated to release a respective bioink.

Systems and methods for in vivo multi-material bioprinting

Described are systems and methods for multi-material printing. The systems and methods can utilize a stereolithographic printing device, a moving stage, and a microfluidic device. The microfluidic device can include a plurality of reservoirs, each reservoir housing a different ink for printing, and a microfluidic chip. The microfluidic chip can include a chamber that comprises a plurality of inlets, a printing region, and one or more outlets as well as an elastic membrane.

Microfluidics-enabled multimaterial stereolithographic printing

Data is collected from a network graph, wherein the collected data is useful for training a machine learning model on a query domain. A domain-specific template corresponding to the query domain is received, the domain-specific template defining one or more classifiers to guide collection of content relevant to the query domain from the network graph. A collection starting point is analyzed based on the one or more classifiers of the domain-specific template to identify one or more relevant instances of the content. The one or more identified relevant instances of the content are added to a contextual protocol package. Each identified relevant instance of the content is analyzed based on the one or more classifiers of the domain-specific template to identify one or more additional relevant instances of the content. The one or more identified additional relevant instances of the content are added to the contextual protocol package.

Guided source collection for a machine learning model

The present disclosure generally relates to systems, methods, and computer-readable media for developing and implementing workflows for a variety of data types. For example, systems disclosed herein may receive or otherwise generate a schema object on a schema system including a plurality of schema objects associated with different workflows. The schema object may include user interface behavior data indicating a content type and associated control type. The schema object may further include application programming interface (API) behavior data indicating a binding between a user interface engine and an API engine. The schema object may also include workflow behavior data indicating one or more services for processing the schema object. Moreover, systems described herein may deploy a plurality of parsers on a plurality of processing engines to enable flexibility and dynamic updates to content ingestion lifecycles.

Ingesting and processing content types

Payloads are received via a communication channel having one-way connectivity. The payloads have a container structure that includes a time stamp. It is determined whether a unit manifest has been received via the communication channel transmitted independently of the payloads. The unit manifest includes information associated with payloads including a hash of each payload on the unit manifest. Responsive to determining that the unit manifest has been received, it is determined whether the information associated with the payloads matches the corresponding payloads. Responsive to determining that the information associated with payloads on the unit manifest matches the corresponding payloads, the corresponding payloads are processed. It is determined whether an aggregate manifest associated with a predefined time period has been received. Responsive to determining that the aggregate manifest has been received, it is determined whether aggregate manifest information from the aggregate manifest matches payloads from the predefined time period.

Data consistency and integrity for one-way connected systems

Solutions for automated labeling of child objects within tagged parents include: receiving a plurality of parent objects, each having a tag and including a plurality of child objects; receiving a machine learning (ML) component operable to rank objects according to relevance to queries; for each parent object: generating a set of restricted objects, wherein each restricted object corresponds to each child object in the plurality of child objects; for each of a plurality of queries, ranking, with the ML component, the restricted objects according to relevance; based at least on the query and an inverse of the rank of the restricted objects, assigning a child object label.

Automated labeling of child objects within tagged parents

A bioreactor and methods for use can include a microfibrous scaffold, that can be made of a composite bioink, and that can have endothelial cells directly embedded within the scaffold using an additive manufacturing process. The scaffold can further be seeded with cardiomyocytes. The hydrogel scaffold can be composed of a plurality of serpentine layers, with each serpentine layers, which can be placed on each other in a cross-hatch configuration, so that the primary axes of successive layers are perpendicular. This configuration can establish an aspect ratio for the scaffold, which can be selectively varied. For greater strength, the successive layers that have a primary axis in the same direction can be placed in the scaffold so that they are slight offset from each other. The scaffold can be placed in the bioreactor with perfusion, for use in cardiovascular drug screening and other nanomedicine endeavors.

Additive manufacturing of functional myocardial tissue

Solutions for automated labeling of child objects within tagged parents include: receiving a plurality of parent objects, each having a tag and including a plurality of child objects; receiving a machine learning (ML) component operable to rank the parent objects; for each parent object: generating a set of restricted objects, wherein each restricted object is missing a child object; for each of a plurality of queries, ranking, with the ML component, the restricted objects according to relevance; based at least on the query and an inverse of the rank of the restricted objects, assigning a child object label. In some examples, the object missing from the restricted object having the lowest relevance rank is assigned a label based on the query. Various advantageous uses of the assigned child labels include enhancing search result reporting, automatically generating ML training data, and automatically generating and/or modifying reports.

A computer-implemented method for training a neural network that is configured to generate a score distribution over a set of multiple output positions. The neural network is configured to process a network input to generate a respective score distribution for each of a plurality of output positions including a respective score for each token in a predetermined set of tokens that includes n-grams of multiple different sizes. Example methods described herein provide trained neural networks which produce results with improved accuracy compared to the state of the art, e.g. translations that are more accurate compared to the state of the art, or more accurate speech recognition compared to the state of the art.

Processing text sequences using neural networks

A speech recognition neural network system includes an encoder neural network and a decoder neural network. The encoder neural network generates an encoded sequence from an input acoustic sequence that represents an utterance. The input acoustic sequence includes a respective acoustic feature representation at each of a plurality of input time steps, the encoded sequence includes a respective encoded representation at each of a plurality of time reduced time steps, and the number of time reduced time steps is less than the number of input time steps. The encoder neural network includes a time reduction subnetwork, a convolutional LSTM subnetwork, and a network in network subnetwork. The decoder neural network receives the encoded sequence and processes the encoded sequence to generate, for each position in an output sequence order, a set of substring scores that includes a respective substring score for each substring in a set of substrings.

Very deep convolutional neural networks for end-to-end speech recognition

Technology related to tools for supporting multi-language software programs is disclosed. In one example of the disclosed technology, a method can be used for testing a user interface of a software product. The method can include receiving a test recording captured using a first build of the software product targeted to a first language. A localized language file corresponding to a second language can be received. The test recording can be played back on a second build of the software product targeted to the second language. The playing back can include finding an equivalent under-test user interface element to a recorded user interface element of the test recording using a hierarchical search of properties in the localized language file. Additionally, the playing back can include performing a recorded action of the test recording on the equivalent under-test user interface element.

Multi-language playback framework

A sequence recognition system comprises a prediction component configured to receive a set of observed features from a signal to be recognized and to output a prediction output indicative of a predicted recognition based on the set of observed features. The sequence recognition system also comprises a classification component configured to receive the prediction output and to output a label indicative of recognition of the signal based on the prediction output.

Prediction-based sequence recognition

Systems and methods for delivering digital content are provided. In one example, a method may include registering a plurality of content source servers and a plurality of content delivery servers with a content management server. The method may include providing a respective inbound content adapter for each of the content source servers, the inbound adapter configured to generate canonical source data. The method may further include providing a respective outbound content adapter for each of the content delivery servers, the outbound adapter configured to process the canonical source data according to a data contract for the respective content delivery server and generate deliverable content data appropriate for the associated content delivery server.

Growing community of inventors

Cambridge, MA, United States of America

Yu Shrike Zhang