---

Neurosymbolic AI

Project Overview

We are seeking motivated researchers (Ph.D., MS, and Undergraduates) to work on cutting-edge research in Neurosymbolic AI, integrating symbolic structure with deep learning to create transparent and interpretable AI systems.

This position focuses on bridging the gap between neural learning and symbolic reasoning, enabling AI systems that are both powerful and explainable.

---

Research Focus Areas

1. Formal World Models and Neural Architectures

   • Representing relational, temporal, and causal structure through formal world models grounded in logic or probabilistic semantics

   • Integrating these structured representations with deep architectures to enable learning under explicit constraints

   • Designing differentiable interfaces that allow neural modules to query, update, and reason over symbolic world states

2. Interpretable Neural Reasoning

   • Developing neural reasoning systems with tractable, polynomial-time inference guarantees

   • Extracting logical rules, program-like structures, or circuit-level decision boundaries from trained networks

   • Building explainability methods that reveal how latent representations support multi-step reasoning, abstraction, and generalization

3. Tractable Neurosymbolic Integration

   • Combining symbolic knowledge bases with tractable probabilistic models such as probabilistic circuits, SDDs, and arithmetic circuits

   • Ensuring efficient exact or approximate inference when coupling discrete logical structure with continuous neural components

   • Creating scalable neurosymbolic pipelines for structured prediction, programmatic reasoning, and out-of-distribution generalization

---

What You’ll Work On

• Designing and implementing new neurosymbolic architectures that couple structured reasoning with high-capacity neural models

• Developing training and inference methods that enforce logical, relational, or probabilistic constraints during learning

• Applying these methods to vision, video understanding, and multimodal reasoning tasks that require grounding in structured world knowledge

• Building interpretable and verifiable models for real-world decision-making settings

• Prototyping algorithms, conducting large-scale experiments, and evaluating systems on challenging benchmarks

---

What We’re Looking For

Essential

• Strong background in machine learning, deep learning, and modern neural architectures
• Proficiency in Python and contemporary DL frameworks (PyTorch, JAX), with an emphasis on efficient, well engineered code
• Solid grounding in algorithms, probability theory, and numerical optimization
• Demonstrated experience conducting reproducible research, including version control, testing, experiment tracking, and documentation

Nice to Have

• Experience with symbolic reasoning, logic programming, constraint solving, or knowledge representation formalisms
• Familiarity with graphical models, probabilistic programming languages, or tractable probabilistic circuits
• Publications in top-tier ML or AI venues (NeurIPS, ICML, AAAI, etc.)

---

What You’ll Gain

• Advanced expertise in neurosymbolic AI, interpretable machine learning, and formal reasoning methods
• Hands-on experience with state-of-the-art deep learning, symbolic reasoning frameworks, and hybrid architectures
• Opportunities to lead and coauthor publications at top ML and AI venues such as NeurIPS, ICML, ICLR, AAAI, etc.

---

How To Apply

Please submit your details using the Google Form.

Note: Choose “Neurosymbolic AI” for the question “Which of the following match your research interests?” in the google form.

Selected students may be invited for a brief meeting to discuss fit and potential directions.

---