Towards Conversational AI for Human-Machine Collaborative MLOps

TL;DR

The paper addresses the accessibility gap in complex MLOps platforms by introducing Swarm Agent, an LLM-driven conversational system that orchestrates Kubeflow Pipelines, data storage, and domain knowledge through specialized agents. It leverages iterative reasoning, tool invocation, and Retrieval-Augmented Generation to enable intuitive pipeline discovery, execution, monitoring, and knowledge integration across users with varying technical backgrounds. The core contributions include the modular Swarm Agent architecture, the KFP, MinIO, and RAG agents, and a concrete implementation that demonstrates end-to-end conversational MLOps workflows with representative use cases. This approach lowers barriers to entry for advanced ML operations while offering extensibility to other platforms and knowledge sources, potentially accelerating collaborative ML development and deployment.

Abstract

This paper presents a Large Language Model (LLM) based conversational agent system designed to enhance human-machine collaboration in Machine Learning Operations (MLOps). We introduce the Swarm Agent, an extensible architecture that integrates specialized agents to create and manage ML workflows through natural language interactions. The system leverages a hierarchical, modular design incorporating a KubeFlow Pipelines (KFP) Agent for ML pipeline orchestration, a MinIO Agent for data management, and a Retrieval-Augmented Generation (RAG) Agent for domain-specific knowledge integration. Through iterative reasoning loops and context-aware processing, the system enables users with varying technical backgrounds to discover, execute, and monitor ML pipelines; manage datasets and artifacts; and access relevant documentation, all via intuitive conversational interfaces. Our approach addresses the accessibility gap in complex MLOps platforms like Kubeflow, making advanced ML tools broadly accessible while maintaining the flexibility to extend to other platforms. The paper describes the architecture, implementation details, and demonstrates how this conversational MLOps assistant reduces complexity and lowers barriers to entry for users across diverse technical skill levels.

Figures (5)

• Figure 1: Architecture of the Swarm Agent system for conversational MLOps.
• Figure 2: Internal Architecture of the KubeFlow Pipelines (KFP) Agent.
• Figure 3: Internal Architecture of the MinIO Agent.
• Figure 4: Internal Architecture of the RAG Agent.
• Figure 5: Architecture of the Knowledge Indexing Pipeline.