About
I'm a Research Engineer at Amazon Web Services in Santa Clara, working on agentic AI. I build agent simulation frameworks that test and validate how AI agents behave in complex, dynamic scenarios, along with LLM-powered systems that redact sensitive information in real time without adding latency.
Before Amazon, I was a graduate researcher at USC's Mark and Mary Stevens Institute of Neuroimaging, studying Tau PET imaging for early detection of Alzheimer's disease. I developed a new partial volume correction algorithm that combined patient MRI and PET scans, running on Apache Spark.
Earlier, at the ITEMS Institute, I built U-Net models for segmenting axons and myelin in microscopy images of nerve tissue.
Writings
Projects
- NeurIPS 2024 BELKA — rank 2 / 1946Kaggle
- Retinal vessel segmentation with GANs
- ADAS image de-hazing with conditional GANs
- ASL fingerspelling recognition with transformers