PhD - Neurosymbolic Reasoning for Domain-Adaptive Generative AI

Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks. However, their performance often degrades in highly specialized domains that require precise structural reasoning, consistency, and domain-specific knowledge. In electronic hardware design, existing LLMs struggle with the structural and semantic complexity of circuit descriptions. Hardware designs are represented as interconnected graphs, including netlists, schematics, and hierarchical design projects. These representations carry essential information that is lost when converted into purely sequential text.

• As a part of our team, you will research novel methods for adapting generative AI systems to specialized domains by integrating structured symbolic knowledge, without relying on extensive model retraining.
• Your investigation will focus on mechanisms that enable Large Language Models (LLMs) and other foundation models to leverage external knowledge representations such as knowledge graphs, ontologies, and formal design representations while preserving the reasoning capabilities of symbolic systems.
• You will develop innovative approaches that enable generative AI models to interact directly with these structured representations, allowing for more precise reasoning, design assistance, verification, and code generation.
• Through your work, you will make a key contribution to the development of next-generation AI systems that combine the flexibility of neural models with the reliability, interpretability, and structured reasoning capabilities of symbolic representations.

Function: Research
Experience Level: Entry Level