Abstract: Systems and methods for estimating a probability distribution of a forecasted quantity includes embedding a training dataset of a trained tree-based model as an array of vectors, each vector corresponding to a respective sample and each dimension representing a decision point in the trained tree-based model. A distance between historical sample points is defined as a cosine similarity function between the vectors. The array of vectors is processed through a Hierarchical Navigable Small World index, and for a new prediction the Hierarchical Navigable Small World index is searched for vectors most similar to a vector associated with the new prediction to create a discrete probability distribution of the forecasted quantity.

Abstract: Systems and methods estimate a probability distribution for a forecasted quantity by embedding a training dataset from a trained tree-based model as vectors, where each sample is a vector and each dimension encodes a decision point in the model. Historical-sample distance is defined using cosine similarity. The vector array is indexed with a Hierarchical Navigable Small World (HNSW) structure; for a new prediction, the index is searched to find vectors most similar to the new prediction's vector, yielding a discrete probability distribution of the forecasted quantity, including quantiles for a product across multiple locations. Tradeoff curves are built for each product-location combination from the quantile predictions. An objective function accepts quantities per location and computes an objective metric from the tradeoff curves.

Abstract: An automated system for generating and maintaining up-to-date architectural diagram codes that effectively combines continuous code monitoring, resource identification, resource annotation, constraint-based diagram generation, and standardized formatting to provide accurate, consistent, and efficient documentation of complex software architectures. This automated approach minimizes manual effort, reduces the risk of errors, and ensures that architectural diagrams remain current with the evolving codebase, thereby enhancing overall change management and communication within development teams.

Abstract: Systems and methods that for a scalable versioned database that can organize data into scenarios and hives. These define how data visibility is controlled by scenarios, and how data can be transferred between scenarios. In some embodiments, queries and algorithms can be executed in independent processes, which may execute in parallel, and on independent machines. Furthermore, data objects can be placed in shared storage, and metadata objects can be placed in a metadata database which supports ACID transactions. Data objects are not modified in place after they are constructed, while metadata objects can be updated transactionally, using the metadata database. Data can be updated by creating new data objects and connecting them to scenarios with new metadata objects. Finally, obsolete data may be cleaned up by identifying unreachable data, which is safe to delete.

Abstract: Systems and methods assess architectural impact of software code changes across multiple repositories. A computing system receives a pull request, identifies and filters irrelevant modifications, and annotates impactful code segments as added, deleted, or modified. The system converts the annotated changes to a graph representation in which software components are represented as nodes and dependencies as edges. The graph representation is compared with a system reference architecture to identify deviations. Based on the deviations and contextual inputs including architectural principles and a current architecture reference, an automated report is generated that indicates components and relationships that are added, removed, or modified and provides an architectural impact assessment. The assessment may be integrated with continuous integration and continuous deployment pipelines to provide commit-level feedback prior to code integration.

Abstract: System and methods for architectural compliance verification by transforming code change annotations into a structured graph representation. The process begins by receiving code change annotations and converting them into a graph where nodes and edges represent code entities and their relationships. The system can filter irrelevant changes and encapsulates impactful modifications with specific tags to isolate significant structural updates. A dynamic change graph can then be constructed and algorithmically compared against a static system architecture reference model. This comparison can identify specific architectural deviations that violate defined principles. Upon detecting non-compliant structural changes, the system can generate a detailed architectural assessment report highlighting the violations.

Abstract: Methods and systems that provide machine learning interpretability. SHAP values of historical and predicted data, along with features of both, are used to provide a measure of the impact of training data points on a predictions. Removal of an individual training data point from a training data set, followed by comparing the resulting prediction with that obtained by the full training data set, also provides a measure of influence of individual training data points on forecasts.

Abstract: Systems and methods that process a settings-related natural-language request in a software platform. This includes receiving, via a user interface, a natural-language request associated with at least one data setting of the software platform and providing the natural-language request to a machine-learning model. Systems and methods further include receiving, from the machine-learning model, a structured output corresponding to the natural-language request and causing the software platform to perform an operation based on the structured output. In some embodiments, the structured output includes a settings-change output identifying a data setting and a setting value. In some embodiments, the structured output is validated against a security policy or a data integrity rule before the operation is performed. The natural-language request may be received through a chat interface and may be generated from voice input using voice-to-text processing.

Abstract: Systems and methods for managing data settings of a software platform includes receiving, via a user interface, a natural-language request from a user. The natural-language request is processed with a routing model to select, from among multiple available workflows, a workflow corresponding to an intended user action. Based on the selected workflow, a corresponding tool is invoked. A machine-learning model generates tool output using the natural-language request and tool-specific instruction content associated with the corresponding tool. The software platform performs the selected workflow based on the tool output and provides, through the user interface, an output corresponding to the performed workflow. In some embodiments, the available workflows include a data settings workflow, a settings summarization workflow, and a resource access workflow.

Abstract: Methods and systems that allow for supply chain logic to be instrumented in such a way that a supply planner can see the major factors driving KPIs, as well as drill down to the lower level to see the impact of each item at the smallest possible level.

Abstract: Systems and methods disclose herein procedures for accelerated tree learning. In one class, the acceleration is based on self-adapting learning rates, while in another class, the acceleration is based on a plurality of learning rates, wherein each learning rate varies over the training; each learning rate increases linearly as a respective pseudo residual maintains a direction across sequential training iterations; and each learning rate decreases exponentially as the respective pseudo residual changes direction across sequential training iterations. The latter can be incorporated with other methodologies, such as momentum-augmented gradient boosting and Nesterov Accelerated Gradient Boosting. These systems and methods for accelerated tree learning exhibit a marked reduction in training time and resources required for gradient boosted trees.

Abstract: Systems and methods disclosed herein provide an automation framework that allows planners to use natural language to explain how they solve problems that they have encountered in the past. This results in an AI agent, conditioned on the instructions provided by a planner. This allows the supply chain management software to recognize problems, produce new scenarios where possible solutions are displayed, and asks the planner to verify which solution is most adequate.

Abstract: Systems and methods for decomposing an optimization problem to reduce solve time. Input data associated with the optimization problem is read and the problem is represented as a graph comprising nodes and edges. A machine learning technique clusters the nodes into a plurality of clusters, and an association between constraints and edges spanning the clusters is obtained. For each constraint, a distribution of predicted load among the clusters is determined based on demand information generated by a machine learning model trained on the input data. At least one constraint is then split across the plurality of clusters in accordance with the predicted load distribution.

Abstract: Systems and methods for features engineering, in which internal and external signals are received and fused. The fusing is based on meta-data of each of the one or more internal signals and each of the one or more external signals. A set of features is generated based on one or more valid combinations that match a transformation input, the transformation forming part of library of transformations. Finally, a set of one or more features is selected from the plurality of features, based on a predictive strength of each feature. The set of selected features can be used to train and select a machine learning model.

Abstract: Systems and methods for editing a file-backed table in a database having a concurrency control mechanism, that include reading the content of the file-backed table, and modifying the content of the file-backed table.

Abstract: Systems and methods for maximizing capacity utilization with constraints, determine a buffer of an already-built (that is, existing) inventory and suggest an increased limit for new inventory that can lead to maximization of capacity. This approach includes the construction of a dynamic ordered buffer list, and a smart search that searches only viable portions of the ordered list for a buffer.

Abstract: Systems and methods for constructing comprehensive test suites for time series machine-learning models are disclosed. The approach leverages frequency-domain analysis to ensure robust coverage and adversarial resilience. A Fast Fourier Transform (FFT) is performed on time series in a test set to extract dominant spectral features, including frequencies, amplitudes, and phases. These features are clustered to group time series with similar temporal patterns, enabling stratified sampling for broad coverage across seasonal and periodic behaviors. Outlier clusters are identified to include rare or edge-case patterns. Additionally, adversarial test cases are generated by perturbing spectral components of sampled time series and reconstructing modified signals via inverse FFT, preserving temporal structure while introducing controlled variations.

Abstract: Methods and systems that allow for supply chain logic to be instrumented in such a way that a supply planner can see the major factors driving KPIs, as well as drill down to the lower level to see the impact of each item at the smallest possible level.

Abstract: Systems and methods for constraint-based optimization, comprising: an AI demand forecasting engine, an optimization engine, a user-defined objective, and a user-defined set of constraints. Using historical sales data, the AI demand forecasting engine generates a plurality of entities, each entity defined by a placement of an item in a promotion platform; and forecasts the objective associated with each entity. The optimization engine generates a plurality of plans, each plan consisting of a unique subset of entities. Plans that violate at least one constraint are eliminated by the optimization engine, leaving a set of candidate solutions. An optimum plan is selected from the set of candidate solutions based on maximization of the objective.

Abstract: A coding assistant for scripting within a user interface of a platform, in which the platform uses customized functions and context. Systems and methods avoid bias (towards generic) functions), verbose and inefficient code, incorrect logic and hallucinations, by using query understanding, information retrieval, code generation, code quality assessment and code revision.