A Deep Dive into AIOps with Selector AI

Selector AI’s presentation at Cloud Field Day 22 offered a technical deep dive into the architecture and functionality of its platform. The core of the platform relies on machine learning (ML) techniques for data ingestion and analysis, ingesting raw data such as metrics, events, and syslogs to understand network behavior without generating content. This ML-driven process focuses on accuracy, utilizing methods like regressions, clustering, and cosine similarities to identify patterns and correlations within the data, avoiding the hallucinations often associated with Large Language Models (LLMs).

The platform’s unique strength is its ability to handle a wide variety of data sources, leveraging a declarative ETL and compiler to easily ingest new data types. This flexibility is showcased by the system’s ability to process data from diverse sources, ranging from legacy network devices and modern cloud services to custom CSV files. The system’s architecture is built on Kubernetes, ensuring horizontal scalability to handle the volume and velocity of data ingested, with a strong focus on creating context through the integration of CMDB data and metadata to give meaning to the raw data. This data integration is a critical component, bringing together disparate data silos to provide a holistic view of network operations.

Generative AI plays a supporting role in Selector AI’s platform, primarily enhancing user experience. It translates natural language queries into the platform’s query language and converts the resulting JSON output back into human-readable English. This use of generative AI is carefully managed to ensure accuracy, acting as a complementary tool to the underlying ML engine and not replacing it. The platform is designed to be flexible, allowing customers to utilize various LLMs and ensuring that the system learns and adapts to the specific details of each customer’s network infrastructure over time. The company emphasizes its commitment to customer support, providing ongoing platform maintenance and support throughout the entire lifecycle of their product implementation.