Abstract: Certain aspects of the present disclosure provide techniques for orchestrating a user experience using natural language input. A user experience is orchestrated within an ecosystem of features for which a plurality of corresponding tokens is defined. Natural language describing a desired user experience result is received by a user experience orchestrator. A sequence of tokens corresponding to operations belonging to an ecosystem of features which produce a correct result for the natural language input can be identified by a trained large language model and executed by the user experience orchestrator using a token operator. The output operations determined by the model to produce or be likely to produce the correct result based on the natural language input can be disambiguated, confirmed, and/or executed.

Abstract: Certain aspects of the present disclosure provide techniques for managing the transmission of mixed-modality messages using machine learning models. An example method generally includes generating, using a first machine learning model, an embedding representation of a mixed-modality message. The mixed-modality message is classified as an effective message or an ineffective message using a second machine learning model and the embedding representation of the mixed-modality message. One or more actions are taken to manage transmission of the mixed-modality message based on the classifying the mixed-modality message as an effective message or an ineffective message.

Abstract: Certain embodiments of the present disclosure provide techniques training a user detection model to identify a user of a software application based on voice recognition. The method generally includes receiving a data set including a plurality of voice interactions with users of a software application. For each respective recording in the data set, a spectrogram representation is generated based on the respective recording. A plurality of voice recognition models are trained. Each of the plurality of voice recognition models is trained based on the spectrogram representation for each of the plurality of voice recordings in the data set. The plurality of voice recognition models are deployed to an interactive voice response system.

Abstract: Certain aspects of the disclosure pertain to inferring a candidate entity associated with a transaction with a machine learning model. An organization identifier and description associated with a transaction can be received as input. In response, an entity embedding, comprising a vector for each entity of an organization based on the organization identifier, can be retrieved from storage. A machine learning model can be invoked with the entity embedding and description. The machine learning model can be trained to infer a transaction embedding from the description and compute a similarity score between the transaction embedding and each vector of the entity embedding. A candidate entity with a similarity score satisfying a threshold can be identified and returned. The candidate entity with the highest similarity score can be identified in certain aspects.

Abstract: Certain aspects of the present disclosure provide techniques for entering user credentials through a proxy. One example method generally includes receiving, at a user device, a push request for user data from a cloud server and receiving a request file from an aggregation system. The method further includes injecting user credentials stored on the user device into the request file, wherein when injected the user credentials replace at least one dummy entry of the request file, and transmitting the request file to a data source associated with the request file. The method further includes receiving user data from the data source and transmitting the user data to the aggregation system.

Abstract: A method includes receiving event strings from source code repositories, creating, for the source code repositories, digests of keywords, receiving log strings, and aggregating the log strings into a log group. The method further includes comparing the digests to the log group to generate scores, whereby the scores correlate the digests to the log group. The method further includes selecting a source code repository from the source code repositories according to the scores, and associating the log group to a service corresponding to the source code repository, where the source code repository corresponds to the digest with a highest score.

Abstract: A method implements anonymous uncensorable cryptographic chains. The method includes receiving, from a first application, verifiable data for a current record and unverified data for the current record. The unverified data for the current record was received by the first application from a second application. The method further includes verifying the verifiable data for the current record with unverified data from a previous record. The method further includes recording the verifiable data for the current record and the unverified data for the current record to the current record responsive to verifying the verifiable data for the current record. The method further includes presenting the current record to one or more of the first application and to the second application.

Abstract: A computer-implemented method includes receiving training data that includes a plurality of API requests from a plurality of client devices. The method includes generating a plurality of permissible API sessions based on the training data. Each of the permissible API sessions is associated with a corresponding client device of the plurality of client devices and includes a sequence of API requests originating from the corresponding client device. The method includes applying a sequence embedding technique to the plurality of permissible API sessions to generate a plurality of embeddings and applying a dimensionality reduction technique to the plurality of embedding to generate a plurality of compact embeddings. The method includes storing each of the compact embeddings in a space partitioning data structure at storage locations within the space partitioning data structure that are determined based on similarities between the compact embeddings.

Abstract: Systems and methods for authenticating a user are disclosed.

Abstract: Certain aspects of the present disclosure provide techniques for training and using predictive models to predict the occurrence of an event within a software application. An example method generally generating a spatially sampled data set for a set of users of a software application. The spatially sampled data set includes, for each respective user of the set of users, an amount of time the user has spent, a number of discrete portions of the software application the user has visited, and an indication of whether the user has completed a defined task. A spatio-temporally sampled data set for users in the spatially sampled data set is generated, including, for each respective user in the spatially sampled data set, a plurality of candidate timestamps. A predictive model is trained based on the spatio-temporally sampled data set.

Abstract: Systems and methods are disclosed for switching between batch processing and real-time processing of time series data, with a system being configured to switch between a batch processing module and a real-time processing module to process time series data. The system includes an orchestration service to indicate when to switch, which may be based on a switching event identified by the orchestration service. In some implementations, the orchestration service identifies a switching event in incoming time series data to be processed. When a batch processing module is to be used to batch process time series data, the real-time processing module may be disabled, with the real-time processing module being enabled when it is used to process the time series data. In some implementations, the real-time processing module includes the same processing models as the batch processing module such that the two modules' outputs have a similar accuracy.

Abstract: A method including receiving a user input for input to a language processing machine learning model (MLM). The method also includes generating modified inputs that are based on, and semantically related to, the user input. The method also includes executing the MLM to generate model outputs of the MLM. The MLM takes as input instances of each of the modified inputs. The method also includes sampling the model outputs using a statistical sampling strategy to generate sampled model outputs. The method also includes clustering the sampled model outputs into clusters. Each cluster of the clusters represents a distinct semantic meaning of the sampled model outputs. The method also includes generating a confidence metric for the user input. The confidence metric includes a predictive entropy of the clusters. The method also includes routing the user input based on whether the confidence metric satisfies or fails to satisfy a threshold value.

Abstract: Certain aspects of the present disclosure provide techniques for predicting activity within a software application using a machine learning model. An example method generally includes generating a multidimensional time-series data set from time-series data associated with activity within a software application. The multidimensional time-series data set generally includes the time-series data organized based on a plurality of time granularities. Using a machine learning model and the generated multidimensional time-series data set, activity within the software application is predicted for one or more time granularities of the plurality of time granularities. Computing resources are allocated to execute operations using the software application based on the predicted activity within the software application.

Abstract: A method comprising generating, during multiple user sessions of a first user with a software application, first clickstream data from the multiple user sessions, and extracting, from the first clickstream data, a first plurality of task instances of the first user performing a first plurality of tasks. The method also includes decomposing, from the first clickstream data, each task instance of the first plurality of task instances into a first plurality of steps to obtain a first plurality of decomposed task instances. The first plurality of steps in the first plurality of decomposed task instances are each associated with a timestamp. The method further includes training a first user model with the first plurality of decomposed task instances to learn a user optimal order to perform the first plurality of tasks and presenting, to the first user, the user optimal order to perform the first plurality of tasks.

Abstract: Certain embodiments of the present disclosure provide techniques for identifying and blocking anomalous transactions within a computing system. An example method generally includes selecting a set of accounts for analysis. For each respective account, an anomaly score is calculated based on an account number associated with the respective account and transaction amounts associated with the respective account. An aggregated anomaly score is generated for each respective account provider of a plurality of account providers based on the anomaly score associated with each respective account. The aggregated anomaly score for each respective account provider is normalized based on a historical minimum and historical maximum anomaly score for the respective account provider. One or more account providers that are potential targets of anomalous activity are identified.

Abstract: Systems and methods for tracking anonymous visitors of an online website or mobile app are disclosed. The browsing activity by an anonymous visitor of the online website or mobile app is converted into features and a visitor-identifier associated with the browsing activity generated by the anonymous visitor is determined. The features are stored with the visitor-identifier in a super-visitor-state before the visitor-identifier is associated with a super-visitor-identifier. After the visitor-identifier is associated with the super-visitor-identifier, the features are stored with the super-visitor-identifier in the super-visitor-state. After the visitor-identifier is associated with the super-visitor-identifier, the features associated with the visitor-identifier in the super-visitor-state may be combined, e.g., aggregated, with the features associated with the super-visitor-identifier and the visitor-identifier may be removed from the super-visitor-state.

Abstract: Certain aspects of the present disclosure provide techniques for privacy preserving sharing and validation of sensitive information in a computing environment. An example method generally includes generating a hashed value of a sensitive data item. A set of modulo values is calculated for the hashed value of the first sensitive data item using a set of prime numbers between an upper bound number and a lower bound number. A request to validate the first sensitive data item is transmitted to a target computing system. The request includes the set of prime numbers and the set of modulo values. An indication of whether a match was found for each respective modulo value in the set of modulo values is received from the target computing system, and a request associated with the first sensitive data item is processed based on the indication.

Abstract: Described herein are example implementations for handling of phishing attempts. A system receives a request to perform an electronic transaction, with the request including information regarding a user account. The system generates one or more probabilities of the request being valid based on the request and processing of a plurality of electronic transactions associated with one or more user accounts, identifies whether the request is valid based on the one or more probabilities, and in response to identifying that the request is not valid, provides an indication that the request is not valid.

Abstract: Certain aspects of the present disclosure provide techniques for generating a user interface to prompt users of a software application to perform an action in the software application. The method generally includes generating historical transaction time gap data for transactions in the account. A probability distribution is generated based on the historical time gap data. The probability distribution represents a probability that a transaction related to the account has been performed after an elapsed time from a previous transaction. A probability that an unrecorded transaction exists for an account based on the probability distribution and a time difference between a most recent transaction and a current time. The probability that an unrecorded transaction exists is determined to exceed a threshold probability, and a user interface is generated and displayed to a user of the software application including a prompt for the user to enter new transactions for the account.

Abstract: Certain aspects of the present disclosure provide techniques for efficient data parity. Embodiments include receiving, by a first data consuming component, from a publication service, a plurality of database change records indicating changes to an underlying data source. Embodiments include updating, by the first data consuming component, a secondary data store based on the plurality of database change records. Embodiments include selecting a sample subset of the plurality of database change records based on one or more conditions. Embodiments include, for each respective database change record in the sample subset that has already been consumed by a second data consuming component that updated a primary data store based on one or more database change records: comparing the respective database change record to first data from the primary data store that corresponds to the respective database change record and making a parity determination based on the comparing.