AutoTuneX: Interactive Automated Fine-Tuning for Large Language Models

We present AutoTuneX, a system architecture design and implementation that lets users interactively fine-tune large language models (LLMs). Our solution is particularly built around Bandit Limited Discrepancy Search [Kishimoto et al., 2022], adapted to perform automated hyperparameter optimization on different LLM tuning strategies. AutoTuneX allows users to upload their data, configure the optimization search space, define and execute tunings, and explore the results of their experiments in an interactive way. Next to the enabling REST API backend, the system comprises of a classical Graphical User Interface (GUI), as well as an Agentic Runtime that lets users operate AutoTuneX conversationally via chat. AutoTuneX enables non-expert users to fine-tune large language models in an engaging way.