Abstract: Confidential tuning of pre-trained machine learning models may be provided. A request associated with a model user account to fine-tune a pre-trained machine learning model with model access restrictions may be received. The pre-trained machine learning model may be one of many pre-trained machine learning models uploaded for selection and fine-tuning. The pre-trained machine learning model may be further trained using a request specified data set, with the model access restrictions and access restrictions for the data set being enforced as part of the training. Then, the fine-tuned machine learning model may be made available for invocation by an application associated with the model user account without violating the model access restrictions and data access restrictions.

Abstract: A connectivity monitor of a vehicle determines current and/or future states of antennas. A workload monitor of the vehicle receives execution criteria for different workloads to be executed. An intelligent connectivity engine at the vehicle receives the current and/or future states of the antennas and the execution criteria for the respective workloads. Based on the current and/or future states of the antennas and the execution criteria for the respective workloads, the intelligent connectivity engine assigns at least one of the respective workloads for current execution and at least another of the respective workloads for future execution. A client may use an intelligent connectivity service to configure various aspects of the vehicle connectivity. For example, the client can provide workload recommendation code for the intelligent connectivity engine to assign workloads for current or future execution.

Abstract: Based on analysis of a workload associated with a throttling key of a client request directed to a first service, a scale-out requirement of the throttling key is obtained at respective resource managers of a plurality of other services which are utilized by the first service to respond to client requests. The resource managers initiate, asynchronously with respect to one another, resource provisioning tasks at each of the other services to fulfill the scale-out requirement. A throttling limit associated with the throttling key is updated to a second throttling key after the resource provisioning tasks are completed by the resource managers, and the updated limit is used to determine whether to accept another client request associated with the throttling key.

Abstract: A decoder is disclosed with error correction for memory data. The decoder's error correction is extended to additional faulty bits by integrating a memory test into the error correction to identify faulty bits in the memory data. A method for correction can include writing a known pattern to the failing address (and possibly to neighboring addresses), reading the known pattern back and comparing the read data to the written pattern to identify the failing bits. The failing bits are then used together with the error correction data to correct memory data having multiple incorrect bits or to alert other components about the failing bit locations.

Abstract: Techniques for dynamically determining a configuration file are described. In an example, a system receives, from a device, a request for first content to be presented by the device. The system determines metadata of the first content and a cue point indicating a start of a segment in which second content can be presented by the device. Based on the metadata and subsequent to the request, the system determines a configuration of the segment. The configuration includes: a duration of the segment, a sequence of slots in the segment, and/or a content duration usable in a slot of the segment. The system generates a configuration file that includes the metadata and the configuration, determines the second content based at least in part on the configuration file, and sends data about the first content and the second content to the device.

Abstract: Devices, systems, and methods are provided for enhanced parsing of manifest files for streaming media. A method may include receiving, by a media player, a first media manifest including tags indicative of first temporal segments for media content; identifying, using a parser, a first tag indicative of a first temporal segment; identifying, using the parser, a second tag indicative of a second temporal segment; parsing, using the parser, the first tag; based on at least one of the parsing or a user preference indicative of a starting location for playback of first media content of the media content or a preferred bandwidth, refraining from parsing the second tag; retrieving, by the at least one processor, based on a time associated with the first temporal segment, an address of the first video content; and presenting, by the at least one processor, the first media content based on the address.

Abstract: This disclosure describes techniques implemented partly by a communications service for identifying and altering undesirable portions of communication data, such as audio data and video data, from a communication session between computing devices. For example, the communications service may monitor the communications session to alter or remove undesirable audio data, such as a dog barking, a doorbell ringing, etc., and/or video data, such as rude gestures, inappropriate facial expressions, etc. The communications service may stream the communication data for the communication session partly through managed servers and analyze the communication data to detect undesirable portions. The communications service may alter or remove the portions of communication data received from a first user device, such as by filtering, refraining from transmitting, or modifying the undesirable portions.

Abstract: A camera device may capture a high-resolution image of a frame and store the high-resolution image in memory. The camera device may down-sample the high-resolution image to a low-resolution image of the frame. The camera device may transmit, to an external compute node, the low-resolution image. The camera device may receive, from the external compute node, a request for a region of interest from the high-resolution image. The camera device may transmit, to the external compute node, the region of interest from the high-resolution image. The camera device may have a captured frame rate at which high-resolution images are captured. The camera device may also have an external frame rate based upon which all, or only some, of the high-resolution images may be down-sampled to low-resolution images that are transmitted to the external compute node. The external frame rate may be decoupled from the captured frame rate.

Abstract: Autonomous mobile robots (AMRs) having adaptive safety systems may operate individually or as part of a convoy. During individual operations, an AMR may determine a safety zone and stop responsive to detecting an object within the safety zone. During convoy operations, an AMR may selectively mute a portion of the safety zone and allow a forward AMR of a convoy within the portion of the safety zone. In this manner, the adaptive safety systems may enable convoy operations of multiple AMRs with relatively greater density and greater speed, thereby improving speed and efficiency of AMR operations without negatively impacting safety.

Abstract: Devices, systems, and methods are provided for rotating cryptography secrets in data hashing. A method may include generating, based on a hashing technique, first obfuscated data using first data and a first version of a secret, and generating first metadata associated with the first obfuscated data, the first metadata indicative of the first version of the secret. The method may include determining a second version of the secret based on the first metadata. The method may include generating, based on the hashing technique, second obfuscated data using the first obfuscated data and the second version of the secret, and generating second metadata associated with the second obfuscated data, the second metadata indicative of the second version of the secret.

Abstract: Intermodal vehicles may be loaded with items and an aerial vehicle, and directed to travel to areas where demand for the items is known or anticipated. The intermodal vehicles may be coupled to locomotives, container ships, road tractors or other vehicles, and equipped with systems for loading one or more items onto the aerial vehicle, and for launching or retrieving the aerial vehicle while the intermodal vehicles are in motion. The areas where the demand is known or anticipated may be identified on any basis, including but not limited to past histories of purchases or deliveries to such areas, or events that are scheduled to occur in such areas. Additionally, intermodal vehicles may be loaded with replacement parts and/or inspection equipment, and configured to conduct repairs, servicing operations or inspections on aerial vehicles within the intermodal vehicles, while the intermodal vehicles are in motion.

Abstract: Customized navigation maps of an area are generated for autonomous vehicles based on a baseline map of the area, transportation systems within the area, and attributes of the autonomous vehicles. The customized navigation maps include a plurality of paths, and two or more of the paths may form an optimal route for performing a task by an autonomous vehicle. Customized navigation maps may be generated for outdoor spaces or indoor spaces, and include specific infrastructure or features on which a specific autonomous vehicle may be configured for travel. Routes may be determined based on access points at destinations such as buildings, and the access points may be manually selected by a user or automatically selected on any basis. The autonomous vehicles may be guided by GPS systems when traveling outdoors, and by imaging devices or other systems when traveling indoors.

Abstract: A post-task-completion retention period for which a computing resource is to be retained, without de-activating the resource, on behalf of a set of requesters of machine learning tasks is determined at a machine learning service. A first task, identified at the service prior to expiration of the retention period at a first computing resource at which a second task has completed, is initiated at the first computing resource. In response to obtaining an indication of a third task and determining that a threshold criterion associated with the retention period satisfies a criterion, the third task is initiated at an additional computing resource. The additional computing resource is de-activated after the third task completes, without waiting for the retention period to expire.

Abstract: An interactive interpretation session with respect to a first version of a machine learning model is initiated. In the session, indications of factors contributing to a prediction decision are provided, as well indications of candidate model enhancement actions. In response to received input, an enhancement action is implemented to obtain a second version of the model. The second version of the model is stored.

Abstract: This disclosure describes, in part, a mobile apparatus that includes capacitive sensor(s) for detecting the presence of a user. For instance, the mobile apparatus, such as a shopping cart, may include capacitive sensor(s) located proximate to a main frame, the handles, and/or the wheel frame. The mobile apparatus may use one or more of these capacitive sensor(s) to determine that the user is located proximate to the mobile apparatus, in contact with the mobile apparatus, and/or located over the main frame of the mobile apparatus. In some examples, the mobile apparatus may then perform one or more actions based on the presence of the user. For example, the mobile apparatus may activate and/or deactivate one or more devices based on the presence of the user. This way, the mobile apparatus is able to conserve power by using the capacitive sensor(s) to detect the present of the user.

Abstract: At a logging service, a representation of a transaction requested by a client is received. The transaction includes at least a first write operation directed to a data object at a first data store. Based at least in part on the contents of the representation and contents of a persistent log of the logging service, a determination is made that the requested transaction is to be committed. A log record corresponding to the first write operation is inserted into the persistent log. The first write is propagated to the first data store after the record has been inserted in the persistent log.

Abstract: Systems, devices, and methods are provided for validation, deletion, and/or recovery of resources in a service environment. A machine (e.g., server) may receive a request to identify or discover a list of resources that are unused in a service environment. A machine (e.g., server) may receive a request to delete one or more resources in a service environment. In at least one embodiment, deletion of a resource involves a two-stage process where the resource is recoverably deleted in a first stage (e.g., by deactivating or disabling the resource) such that the resource can be recovered prior to a predetermined time period by reactivating or re-enabling the resource and, in a second stage, the resource is unrecoverably deleted.

Abstract: Systems and methods are described herein for repairing a block of pixels in image data that was encoding using a lossy type of compression. In one aspect, it may be determined that a first pixel block within a first frame of image data was encoded using lossy compression, where the first pixel block comprising a plurality of pixels. Next, it may be determined that the plurality of pixels in the first pixel block in a second frame of the image data are below a threshold difference as compared to the plurality of pixels within the first frame of the image data. Responsive to the determining, the first pixel block may be re-encoded in the second frame or a subsequent frame of the image data using lossless compression.