A
A — AI
Resource-Efficient Foundation Models for Code
Quantization, parameter-efficient fine-tuning, and distillation that make code-intelligence models cheap enough to deploy at developer scale — without sacrificing quality.
Research focus
Four threads spelling A·U·R·A.
AI for code · Understanding what models do · Reliable evaluation of how they behave · Automation that practitioners can trust. Each letter names a multi-paper research line we run at the AURA Lab.
U
U — Understanding
Neurosymbolic Reasoning & Interpretability
Combining neural language models with grammars, type systems, and program analysis — plus feature-level explanations that move beyond opaque token-by-token predictions.
R
R — Reliable
Reliability, Causality & Honest Evaluation
Causal reasoning, counterfactual analysis, and rigorous LLM-as-a-Judge methodologies that measure what matters in code AI — not what's easy to measure.
A
A — Automation
AI Agents for the Software Lifecycle
Designing autonomous agents that plan, reason, and execute multi-step software workflows — from issue triage to code review — with measurable end-to-end behavior.