The Symphony AyasdiAI platform uses Apache Hadoop and Spark to analyze enterprise-scale data to generate machine learning optimized features. The platform supports customer data from enterprise data stores, including relational databases, HDFS files, CSV, HIVE, ranging from analyst’s local files to large enterprise clouds. The integrated Workbench data science tool enables users to create topological models of their data and visually explore the relationships inherent in those models. Architected with Java services and Django app servers, the REST API and Python SDK allow for rapid analysis, development, and expansion of solutions and integrations. The web-based Envision development framework accelerates development through a collaborative, workflow-oriented approach that includes business users throughout the process, and a Python SDK for programmatic access to the technology layer. Symphony AyasdiAI solutions are built on React HTML technology, giving a responsive and portable user experience.

Symphony AyasdiAI stands alone in providing AI solutions techniques that make machine learning models and their decisions interpretable, in contrast with “black box” machine learning methods. Explainability is important to ensure that rules created by machine learning algorithms are objective. Explainable AI is critically important in heavily regulated industries like financial services where the ability to explain decisions made by AI is equally important as the accuracy of the decisions themselves.

Real-world data often comes at enormous volume and high dimensionality. In order to derive concrete insights, the sheer volume of this data often requires statistical methods such as data sampling to render the data digestible for humans, with the undesirable side effect of a significant loss of information. In contrast, the Symphony AyasdiAI patented technology makes this difficult task uniquely feasible via intuitive visualizations and statistical aids that retain all essential attributes of the data, unleashing the full potential of the data to discover key patterns and changes in key business metrics that affect the bottom line.

Data scientists can use Symphony AyasdiAI’s TDA to visualize an entire dataset while keeping all relevant characteristics of the data. The “shape” of the data can help validate AI model predictions (classifications) by comparing characteristics of identified data subpopulations with the true label of that prediction. This is useful when validating and identifying errors in classification AI models. Analysis of these prediction groups can help explain the reasoning of an AI model, by understanding the behavior characteristics of that group. Data scientists can use these insights to improve model performance by correcting errors in model assumptions.

Enterprise AI solutions require handling vast amounts of multi-dimensional data from diverse sources. Our platform deals with the complexities and scale of virtually any data store with ease, performing equally well with hundreds of concurrent users, across hundreds of node clusters, and handling real-time data as a deployable microservice. One global banking project involved ingesting and analyzing 600 million lines of logs from six data centers, amounting to 3 TB of data every month, to discover patterns and develop predictive analytics. Another global bank used our solution to analyze the data of 18 million customers, across 200 million transactions—a full year’s worth of data—to discover potential fraud.

With the Symphony AyasdiAI platform, developers can speed through the development lifecycle of a deployable AI application, where all it’s component aspects such as integrating with existing data, applying repeatable modeling processes, creating appropriate user interfaces, setting up and integrating with enterprise-level authentication, security, and permissions settings within a week. Data ingestion is a snap, the platform connects to any virtual data source—Postgres, SQL, Oracle, SAP Hana, SalesForce, Aurora, Cassandra, S3, Hadoop, DynamoDB, MongoDB, or anything else. Our platform uses TDA to discover the schema of your data, freeing developers, and business analysts from the time-consuming task.

The Symphony AyasdiAI platform leverages TDA under the hood and exposes its discoveries rapidly through the application of unsupervised learning and machine learning algorithms. Add any new data and within a matter of minutes, insights, patterns, and testable hypotheses, along with their statistical significance begin to appear from the data. One large pharmaceutical company attested that it took two days to discover novel associations in clinical response findings that would have taken two years without the Symphony AyasdiAI platform.