Abstract: Systems and processes are described for a message service with distributed key caching for server-side encryption. Message requests are received by message handlers of the message service that cache data encryption keys used to encrypt and decrypt messages that are stored to message containers in back end storage. A metadata service obtains the data encryption keys from a key management service, caches the keys locally, and sends the keys to the message handlers upon request, where the keys are cached, again. The key management service may generate the data encryption keys based on a master key (e.g., a client's master key). The message handlers may send both message data encrypted using the data encryption key and an encrypted copy of the data encryption key to be stored together in the data store.

Abstract: Devices and techniques are generally described for generating response templates for natural language processing. In various examples, a first knowledge graph comprising a plurality of entities may be received. First text data may be received for a first response template, the first text data defining a natural language input configured to invoke the first response template. A response definition may be received for the first response template, the response definition defining a response associated with the first response template. Natural language input data may be received. A determination may be made that the natural language input data corresponds to the natural language input configured to invoke the first response template. The first response template may be configured to generate natural language output data.

Abstract: Techniques for monitoring devices to use ultrasonic signals to detect and track the locations of moving objects in an environment. To determine distance information, the monitoring devices emit a frequency-modulated continuous wave (FMCW) signal at an ultrasound frequency range. Reflections of the FMCW ultrasonic signal are used to generate time-of-arrival (TOA) profiles that indicate distances between the monitoring device and objects in the environment. The reflections can be processed to suppress undesirable interferences, such as reflections off non-mobile objects in the environment (e.g., walls, furniture, etc.), vibrations off the floorings or the ceilings, etc. After processing the reflections, a heatmap can be used to plot the intensity of the reflections for the different TOAs of the reflections, and depict the movement of the user over time. Finally, a Kalman filter is used to smooth the peaks in the intensity values on the plot, and determine the trajectory of the human.

Abstract: A method and apparatus for testing and simulating an access control policy are disclosed. Evaluating an access control policy may be performed by utilizing a deny statement that causes the access request to be rejected despite actions indicated in the access request being authorized. Further, an independent simulation environment may be utilized for testing access control policy evaluation.

Abstract: A semiconductor package can include a unitary substrate defining a first region and a second region. A first interposer can be mounted on the unitary substrate in the first region, and a first die can be mounted on the first interposer. A second interposer can be mounted in the second region on the unitary substrate, and a second die can be mounted on the second interposer.

Abstract: Joint noise and echo suppression may be performed for enhancing two-way audio communications. Audio data is captured at a communication device and audio data transmitted to the communication device from another communication device are used as input features to a trained machine learning model that uses the transmitted audio data as a reference signal to eliminate residual echo in the captured audio data when also suppressing noise in the captured audio data.

Abstract: At a first compute instance run on a virtualization host, a local instance scaling manager is launched. The scaling manager determines, based on metrics collected at the host, that a triggering condition for redistributing one or more types of resources of the first compute instance has been met. The scaling manager causes virtualization management components to allocate a subset of the first compute instance's resources to a second compute instance at the host.

Abstract: Where an event is determined to have occurred at a location within a vicinity of a plurality of actors, imaging data captured using cameras having the location is processed using one or more machine learning systems or techniques operating on the cameras to determine which of the actors is most likely associated with the event. For each relevant pixel of each image captured by a camera, the camera returns a set of vectors extending to pixels of body parts of actors who are most likely to have been involved with an event occurring at the relevant pixel, along with a measure of confidence in the respective vectors. A server receives the vectors from the cameras, determines which of the images depicted the event in a favorable view, based at least in part on the quality of such images, and selects one of the actors as associated with the event accordingly.

Abstract: When a client requests a data import job, a remote storage service provider provisions a shippable storage device that will be used to transfer client data from the client to the service provider for import. The service provider generates security information for the data import job, provisions the shippable storage device with the security information, and sends the shippable storage device to the client. The service provider also sends client-keys to the client, separate from the shippable storage device (e.g., via a network). The client receives the device, encrypts the client data and keys, transfers the encrypted data and keys onto the device, and ships it back to the service provider. The remote storage service provider authenticates the storage device, decrypts client-generated keys using the client-keys stored at the storage service provider, decrypts the data using the decrypted client-side generated keys, and imports the decrypted data.

Abstract: A multi-assistant speech-processing system that centrally determines multiple execution plans to respond to a user input. A central component determines whether a particular input should be processed using a requested assistant or a different assistant or should be terminated. Assistant handoff may be determined based on system policies as well as user input-specific data. A ranked list of execution options may be supplemented by augmented data corresponding to messages to a user. The system may attempt to execute plans in the ranked order until a plan succeeds.

Abstract: Devices and techniques are generally described for iterative theme discovery in text. In some examples, first text data that is separated into a plurality of documents may be received. In some examples, a codebook comprising a first topic associated with a first set of keywords and a second topic associated with second set of keywords may be identified. Instructions to modify the codebook to generate a modified codebook may be received. The instructions may be effective to add to, delete from, and/or modify at least one of the keywords of the first set of keywords or the first topic. In some further examples, a first document of the plurality of documents may be tagged with the first topic based at least in part on the first keywords of the modified codebook. In some examples, output data may be generated that indicates that the first document pertains to the first topic.

Abstract: Techniques for classifying and processing physical objects are disclosed. In an example, a computer system may receive first data indicating that a first machine learning model of a robotic system is incapable to classify a physical object according to at least one of a set of predetermined classifications. The computer system may also receive second data corresponding to one or more attributes associated with the physical object. A second machine learning model of the computer system may determine a cluster of physical objects that includes an identifier of the physical object, whereby the identifier is included in the cluster based at least in part on the first data and a common attribute with other physical objects of the cluster. The computer system may then determine data for processing subsequent physical objects that are determined to have the common attribute.

Abstract: Commercial interactions with non-discretized items such as liquids in carafes or other dispensers are detected and associated with actors using images captured by one or more digital cameras including the carafes or dispensers within their fields of view. The images are processed to detect body parts of actors and other aspects therein, and to not only determine that a commercial interaction has occurred but also identify an actor that performed the commercial interaction. Based on information or data determined from such images, movements of body parts associated with raising, lowering or rotating one or more carafes or other dispensers may be detected, and a commercial interaction involving such carafes or dispensers may be detected and associated with a specific actor accordingly.

Abstract: A pair of virtualized security device initialization data sets are received at a first virtualization server from respective sources prior to a launch of a compute instance at the server. A first virtualized security device is initialized using the data sets, and used to generate cryptographic artifacts used by the compute instance. A data item which was included in one of the data sets is modified after the cryptographic artifacts are generated. Additional cryptographic artifacts are generated by a second virtualized security device at a second virtualization server to which the compute instance has been migrated. The second virtualized security device is initialized using at least the modified data item.

Abstract: Systems, methods, and computer-readable media are disclosed for flexible rail switching components for container shuttles. In one embodiment, an example system for a shuttle may include an elongated spring steel plate, a first rail track component coupled to the elongated spring steel plate, a second rail track component coupled to the elongated spring steel plate and disposed adjacent to the first rail track component, and a third rail track component coupled to the elongated spring steel plate and disposed adjacent to the second rail track component. The elongated spring steel plate may be configured to be actuated from a first position to a second position. A distance between the first rail track component and the second rail track component when the elongated spring steel plate is in the second position may be less than when the elongated spring steel plate is in the first position.

Abstract: A safety rated robotic motor control system may be configured to receive sensor data from sensors positioned on a motor. The data can be received by a programmable logic device and processed to generate a fault message based on a first set of characteristics. The fault message and the data can be transmitted to a safety controller that determines a second set of characteristics and generates operating instructions based on the fault message and the second set of characteristics.

Abstract: Disclosed are systems and methods that determine specification portions of Dafny code and transform those specifications into one or more annotations, expressions, comments, and/or assertions that are included in a destination code written in a destination language as part of a compilation of the Dafny code into the destination code. The annotations, expressions, comments, and/or assertions in the destination code may be utilized by a verification component, such as a Checker Framework, to detect errors that are introduced into the destination code by the compiler as part of the compilation or to verify the absence of errors in the destination code.

Abstract: Techniques are generally described for fusing sensor data of different modalities using a transformer. In various examples, first sensor data may be received from a first sensor and second sensor data may be received from a second sensor. A first feature representation of the first sensor data may be generated using a first machine learning model and a second feature representation of the second sensor data may be generated using a second machine learning model. In some examples, a modified first feature representation of the first sensor data may be generated based at least in part on a self-attention mechanism of a transformer encoder. The modified first feature representation may be generated based at least in part on the first feature representation and the second feature representation. A computer vision task may be performed using the modified first feature representation.