Research
Research Interests
The directions I'm actively studying and building toward. These are declared interests — areas where I'm developing depth through coursework, paper reproductions, and personal projects — rather than an established publication record (yet).
Robust & Trustworthy AI
Building ML systems whose behavior holds up under distribution shift and adversarial conditions — an interest grounded in hands-on work on drift detection and ML monitoring.
Related work
Neural Network Verification
Formal methods for proving properties of neural networks: reachability analysis, abstract interpretation, and certified robustness bounds.
Formal Verification
Mathematical techniques for guaranteeing system correctness, and how they can be extended from classical software to learned components.
Computer Vision
Modern vision architectures, especially transformer-based models — explored through a from-scratch ViT paper reproduction in PyTorch.
Related work
Language Models & Retrieval
How LLMs ground their answers in external knowledge: RAG pipelines, vector retrieval, and protocol-level integrations like MCP.
Related work
Optimization & Learning Theory
The mathematics underneath training: Bayesian hyperparameter optimization, learning-rate schedules, and statistical evaluation of models.
Related work