Late Breaking Results: ‘Describing’ Your Way to a Functional Microfluidic Chip

Microfluidic devices enable miniaturized, automated laboratory operations, but their design typically requires specialized expertise and labor-intensive CAD workflows. Translating an experimental goal into a manufacturable chip layout remains a highly specialized and iterative process, leaving a semantic gap between high-level intent and machine-readable design descriptions.

We present a language-driven framework that synthesizes manufacturable microfluidic designs directly from natural-language prompts. The framework introduces an LLM-based high-level synthesis paradigm, using domain-specific fine-tuned models (Llama 3.3) to translate user intent into a structured JSON design specification for subsequent layout synthesis.

To address the lack of domain-specific training data, we introduce a scalable synthetic data generation pipeline that programmatically creates diverse, valid microfluidic topologies and produces multiple prompt styles with LLM-based linguistic augmentation. An automated validation and correction pipeline detects and fixes geometric and physical inconsistencies—including JSON schema violations and topological errors—to ensure design viability. Validated designs are then synthesized into 2D and 3D geometries for conventional and 3D-printing fabrication, enabling end users to generate and refine microfluidic designs through a conversational interface.