Nexla Architecture
Converged integration built on data products, powered by AI.
AI-Powered Integration Built on Nexsets
Powered by AI
Nexla is built from the ground up to let any data users find, access, integrate, govern, and share data without having to integrate first. Point and click within the Nexla Designer or type prompts in the Nexla Orchestrated Versatile Agent (NOVA) interface, browse Nexla’s built-in marketplace, and find the right data. Nexla does the rest using continuous metadata intelligence, powered by AI, so you can focus on your data.
Nexla is built on Nexsets, virtual data products that turn any structured or unstructured data into human-readable, ready-to-use data independent of formats, protocols, or update speeds. Just connect to any source using Nexla’s universal bidirectional connectors, or use any API. Nexla automatically turns sources into Nexsets, generates data validation rules to improve data quality, and lets you publish and govern Nexsets in your marketplace.
Nexsets make integration simple and fast without coding. Choose any integration style – ETL/ELT/R-ETL, streaming, API Integration, (S)FTP, agentic RAG, or your own custom integration style. Add the Nexsets you need. Transform, merge, add validation rules and your own logic. Finally, load the data into your destinations of choice. Nexla automatically deploys each flow, optimizing it for performance, scalability, and efficiency.
Instant Data Products - Nexsets
The core foundation of Nexla are Nexsets, human-readable, virtual data products that turn heterogeneous data silos into a unified data model. Nexsets make it easy for anyone to find and use data that is validated, documented, discoverable, governed with access and usage controls, and independent of formats, protocols, and data speeds.
Nexla automatically creates Nexsets from any structured or unstructured data sources. All you have to do is connect. Nexla then scans for any data and metadata, and uses its metadata intelligence to prepare and transform it into ready-to-use Nexsets. Nexla automatically creates data validation rules as part of its ingestion pipeline to help transform and ensure data quality. You can add your own rules to automate error correction or responses to exceptions.
Nexsets are the building blocks of all flows as well. You can build any pipeline by endlessly transforming and composing Nexsets into new Nexsets through no-code operations.
Powered by AI - Metadata Intelligence
At the core of Nexla’s automation is metadata intelligence, which uses AI and ML to create and maintain Nexsets, flows, and other Nexla entities. Metadata intelligence is always active, continuously working as data moves. When Nexla connects to a source for the first time it analyzes metadata and record-level data, and auto-generates a ready-to-use, fully governed Nexset with schema, samples, descriptions, annotations, ratings, and other metadata. Nexla also creates validation rules to help automate schema evolution and improve data quality.
Metadata intelligence takes most of the work out of integration, transforming heterogeneous data silos into a unified data model. It turns any source format – unstructured documents, CSV, JSON, XML, Parquet, emails, images – into Nexsets that are independent of formats, protocols, or speeds. As you build pipelines out of Nexsets, Nexla automatically manages all the transformation and mapping.
Automated Connectivity - Universal Connector Architecture
Nexla provides a universal connector architecture to connect to any data sources or system, and convert data into Nexsets. There are 100s of universal bidirectional connectors for SaaS and on-premises apps, APIs, databases, data warehouses, files, file-based storage systems, e-mails, and more. But setting up a new pre-configured connector, or integrating to any API is really fast. Once done, Nexla’s metadata intelligence does the rest.
Easy Collaboration and Governance - Marketplace
The Nexla Marketplace is the central place to discover, share, and govern Nexsets. Consumers easily discover, learn more about, and request access to Nexsets based on their privileges via a simple shopping-like interface. Once you’ve created a Nexset, you can publish it to a marketplace with a few clicks, and share the Nexset with other individuals or teams.
Each marketplace is private to your account, tailored by your administrators to fit each organization’s needs, manage domains, approve publication, and govern all access and usage based on policies.
Converged Integration - Data Flows
The data processing workhorse in Nexla is the data flow. There are different flow types for each integration style. For example, a FlexFlow can mix any integration styles. Specific flow templates like ELT, CDC, RAG, replication, and Spark ETL preconfigure many of the steps of a specific type of flow to make implementation much faster. Once you’ve designed a flow, Nexla automatically provisions each flow type to a specific engine and optimizes based on the source(s), destination(s), latency, and other details.
Flows use Nexsets to automatically extract form and provision data to any sources and applications, databases, data warehouses, data lakes, files, and APIs. Data flows handle any-to-any integration in the truest sense. They allow you to share the same structured and unstructured data in many different formats across operations, analytics, and AI.
Agentic AI - RAG Flows Converged Integration - Data Flows
Nexla combines the power of flows and Nexsets with the latest in agentic AI to help companies deliver accurate, enterprise-grade AI fast. The RAG flow lets you implement end-to-end enterprise-grade agentic RAG without coding.
Nexsets give you easy access to any unstructured or structured data, making sure you have all the data you need. Choose from leading chunking, embedding, and vector databases, including Nexla’s own agentic chunking, to prepare and continuously update data for training or RAG.
Then configure RAG flows on a common modular RAG architecture that helps you improve AI accuracy and lower costs. Choose different LLMs and modules for query rewriting, agentic retrieval, re-ranking, multi-LLM comparison, grading. Use orchestration to create agentic pipelines.
Modern DataOps - Schema Evolution
Continuous metadata intelligence is also used to continuously check and update Nexsets, flows, transforms, and other entities. Validation rules detect schema changes as they occur. Nexla then determines whether to automatically propagate changes from sources to destinations by appending attributes or reversioning, and when to require manual intervention. All changes across connectors, Nexsets, transforms, and flows are automatically versioned and tracked. Built-in lineage and audit trails make it easy to know what changed, who changed it, and when intervention is needed.
Automated Monitoring and Alerting - Data Monitoring
Nexla monitors data access and usage continuously across pipelines. It logs activity from each step including data validation rules, and allows you to add annotations. You can subscribe to receive alerts, and automate responses to errors as issues are detected. Nexla gives you full control over how to handle each error, with built-in lineage and audit trails that let you trace back errors and fix them without disrupting your operations. Nexla also detects data quality and schema issues, letting you quarantine data and either automate data quality or schema evolution, or intervene manually when needed.
Deploy Anywhere – Hybrid Architecture
Nexla is built using a centralized control plane that can support multiple distributed data planes. You can deploy as SaaS in AWS, GCP, and Azure, as hybrid multi-cloud, in a private cloud using VPC, or completely on premises in a data center. You can also integrate Nexla into your environment to help ensure security and governance, and reduce administrative time and costs. Nexla supports single sign-on using 3rd party identity and access management (IAM) tools, automated versioning, end-to-end audit trails, and record-level data lineage.
