Scholar
Swarat Chaudhuri
Google Scholar ID: 9j6RBYQAAAAJ
UT Austin, Google DeepMind
Citations & Impact
Citations
5,381
H-index
36
i10-index
67
Publications
20
Co-authors
76
list available
Contact
Publications
Resume (English only)
Academic Achievements
- Named a 2025 Guggenheim Fellow
- Developed neurosymbolic program synthesis algorithms including sketch learning (Bayou), neural admissible heuristic search (Near), imitation-projected descent (Propel, PIRL), and LLM-aided evolution and abstraction (LaSR)
- Pioneered neurosymbolic program representations such as modular neural architectures (Houdini), compositions of neural predictors and traditional software (DSE), and differentiable relaxations of symbolic programs (Near)
- Recent publications on LLM agents for theorem proving (Copra), search-based natural language deduction (SCSearch), and compositional world modeling (Cosmos)
- Delivered tutorials on neurosymbolic programming at NeurIPS 2022 and POPL 2023
- Co-authored an op-ed on the 2023 AI Safety Executive Order
Research Experience
- Leads the Trishul lab, investigating problems at the interface of automated reasoning, machine learning, and programming languages
- Has worked on program synthesis for many years—early work based on symbolic formal methods (e.g., Lambda2, ConSynth), recent work on neurosymbolic approaches
- Extensive background in automated reasoning: PhD work on automata-theoretic formal verification (Nested Trees); current work on neurosymbolic formal and informal (visual/textual) reasoning
- On leave at Google DeepMind's Science and Strategic Initiatives Unit in London since Fall 2024
- Served as Program Co-Chair for ICLR 2024
