About

Background, current focus, and how to get in touch.

§ 01 · Background

Background

I build deep learning systems for drug discovery and medical imaging — multimodal models, segmentation tools, and the data infrastructure they run on. Since 2019 I’ve been at AstraZeneca R&D in Maryland.

My work spans three threads. Drug discovery ML — multimodal models that fuse cell painting microscopy with SMILES chemical representations to predict compound activity (IC50), with companion graph-based pipelines for multiplex immunofluorescence analysis. Medical imaging and segmentation — transformer-backed interactive segmentation tools for 3D volumetric data, patch-based unsupervised pipelines for imaging at scales that exceed GPU memory. Tooling and infrastructure — a unified ingestion and preprocessing platform that has become the team’s standard data layer (sixteen multi-center datasets, 150K CT volumes), and the visualization suites that R&D labs use to interpret model outputs.

Before AstraZeneca, I was a Software Engineer at Ann Arbor Algorithms, building containerized end-to-end deep learning pipelines for medical imaging — classification, 3D bounding-box detection, anomaly identification — on multimodal datasets of millions of points. I trained as a bioinformatician at the University of Michigan (M.S., 2018) after my biomedical sciences B.S. at the University of Macau.

§ 02 · Now

What I am working on

Recent threads at AstraZeneca

  • Automatic contrast phase classification of polyphasic CT scans (poster, AACR 2026).
  • Multimodal IC50 prediction at scale across compound libraries.

Currently

  • Agentic LLM systems — hand-rolled tool-calling agent loops with pluggable model providers; locally hosted models via Ollama as the primary inference path. Most recent: an agent that autonomously researches stocks across eleven tools, with workflow pipelines included as comparison voices.
  • GPU compute environments — Proxmox/KVM virtualization with PCIe passthrough for specialized ML and cross-platform development workloads.

§ 03 · Skills

Skills

Programming — Python, C++, R.

Machine learning and AI — PyTorch, Keras, scikit-learn, Hugging Face, HDBSCAN, UMAP, MLX, Ollama.

Applied AI / LLM — LLM tool-calling & agent loops, RAG (AnythingLLM), structured outputs, model evaluation, prompt engineering, Ollama / local inference.

Medical AI and imaging — MONAI, nnU-Net, DICOM, NIfTI, ITK / SimpleITK.

Infrastructure and MLOps — Docker, Kubernetes, Git, Proxmox, KVM, Dash, Agile.

§ 04 · Contact

Contact

I’m currently open to new ML / applied-AI opportunities. The best way to reach me is via LinkedIn.

  • GitHub for code and personal projects.
  • For email and phone, see my CV.