Anthony Vizcaíno Aportela

Physics Ph.D. candidate at UC San Diego. My favorite work sits between teaching physics and scientific computing: learning hard ideas clearly, then building the AI, data, and cluster systems that let other people use them.

I build ML models, ROOT-processing workflows, and cluster infrastructure for CMS muon-system shower searches. I care about science IT for the same reason I care about teaching: the work is best when it helps people understand difficult systems and actually get useful work done. Outside formal research, projects like the homelab, ceramics, and small builds are ways to learn new tools by doing concrete work.

• 📬 [email protected]
• GitHub: AnthonyAportela
• ORCID: 0000-0001-9171-1972
• LinkedIn: Anthony Aportela
• 🏠 San Diego, CA
• 📞 (404) 934-0287

Undergraduate Project Advisor

UC San Diego | La Jolla, CA | Oct 2023-Present

• Mentor about 10 undergraduate researchers working on DGCNN-based particle classification and hls4ml deployment to Xilinx U55C FPGAs on NRP.
• Help students get onto UAF, NRP/Nautilus, LXPLUS, and LPC computing environments; review code, set milestones, debug workflows, and support grant applications.

CMS Machine Learning Hands-on Tutorial

Fermilab | Batavia, IL | July 2023

• Taught a hands-on session covering Python, XGBoost, Keras, TensorFlow, PyTorch, graph neural networks, VAEs, and GANs; supported event-classification exercises on the Purdue Analysis Facility and Vanderbilt JupyterHub.

Physics Instruction

UC San Diego and San Diego Mesa College | San Diego, CA | Sep 2020-Dec 2025

• Instructor of record for UC San Diego PHYS 1B in three 2025 terms, with enrollments up to 450 students; also taught two semesters as a Mesa College adjunct professor and six terms as a UC San Diego teaching assistant.
• Designed lectures, laboratories, assessments, and support materials and coordinated teaching assistants and readers across lecture and laboratory courses.

CMS Muon-System Shower Workflows

Lead developer and operator | NRP/Nautilus | 2021-Present

• Ran an Indexed Kubernetes Job with about 50,000 pod completions, capped at 200 concurrent pods and 2-3 analyzer workers per pod; processed nearly one million ROOT files and tens of TB in hours.
• Set up PyTorch DDP jobs across nodes, pods, and GPUs with fixed data plans, checkpoint/restart, MLflow tracking, digest-pinned containers, S3-compatible object storage, and Ceph volumes.
• Wrote the production-run tooling around GitLab, containers, CVMFS, PVCs, and object storage: transfer queues, progress estimates, validation, and targeted retries. Verified 19,680 output objects by exact manifest comparison with zero missing or extra files.
• Debugged NCCL shared-memory and S3 synchronization failures, then restarted eight-rank and long-running DGCNN jobs from durable checkpoints.

Personal Homelab: Infrastructure Practice

Independent infrastructure project | San Diego, CA | 2026-Present

• Built and operated a heterogeneous Proxmox/PXVIRT homelab across x86-64 and ARM hosts, working through VM/LXC lifecycle, quorum-aware placement, ordered startup, and GPU/NVMe passthrough.
• Provisioned native Proxmox/Ceph storage: Squid packages, MON/MGR/MDS services, six HDD OSDs, CephFS, replicated metadata/default pools, and an EC 3+3 bulk-data pool; validated placement, smoke data, larger test corpus reads, and paired-OSD diagnostics before admitting protected data.
• Host and operate apovisc.dev directly from the homelab cluster: GitHub builds the Astro site, publishes a static deploy branch, and an nginx origin guest pulls and serves the release.
• Built an 11-disk MooseFS shared-storage tier with separate master, chunkserver, metalogger, and FUSE-client roles; learned to debug CRC and I/O faults across disk, USB, container, filesystem, and distributed-storage layers.
• Set up private service networking with Tailscale, CoreDNS, and Nginx, plus Ansible/Python automation with schema validation, read-only drift audits, and end-to-end service checks.

Graduate Researcher, ML for CMS Muon-System Showers

UC San Diego Department of Physics and CMS Collaboration | San Diego, CA | Jul 2021-Present

• Train and compare BDT, DNN, Dynamic Graph CNN, and transformer classifiers for long-lived particle decays producing showers in the CMS muon system. A completed DGCNN run reached 0.974 validation AUC; current work tests FlashAttention for the transformer path.
• Keep dataset, split, weighting, checkpoint, inference, and provenance rules fixed so model comparisons are reproducible across distributed training runs.

UC San Diego

La Jolla, CA

• Physics Ph.D., expected Aug 2027 (3.72 GPA) | Oct 2020-Aug 2027
• Physics M.S. | 2023
• Working dissertation title: Machine Learning Methods for Searches for Long-Lived Particles Producing Showers in the CMS Muon System
• Advisor: Javier Duarte
• Selected coursework: Statistics, Parallel Computing, Computational Physics

Georgia Institute of Technology

Atlanta, GA

• Physics B.S. Highest Honors (3.75 GPA) | 2015-2019

Cluster and infrastructure: Kubernetes, Linux, Proxmox/PXVIRT, LXC, Ansible, MooseFS, Ceph/CephFS, CoreDNS, Tailscale, Nginx, GitLab CI/CD, Docker/Podman, Apptainer, Bash, Git, S3-compatible object storage, CVMFS

Machine learning and scientific data: Python, PyTorch, PyTorch Geometric, PyTorch DDP/torchrun, NCCL, MLflow, XGBoost, NumPy, SciPy, Pandas, Matplotlib, ROOT, pyROOT, uproot, Awkward Array, coffea

Accelerators and FPGAs: CUDA, FlashAttention, hls4ml, AMD/Xilinx Alveo U55C, C++, C, written/conversational Spanish

• E. Chien, M. Li, A. Aportela, et al. “Opportunities and challenges of graph neural networks in electrical engineering.” Nature Reviews Electrical Engineering, 2024.
• A. McCarn Deiana et al., including A. Aportela. “Applications and techniques for fast machine learning in science.” Frontiers in Big Data, 2022.
• S. Tsan, R. Kansal, A. Aportela, et al. “Particle Graph Autoencoders and Differentiable, Learned Energy Mover’s Distance.” NeurIPS ML4PS Workshop, 2021.

Presentations

• Enhancing the Search for Long-Lived Particles in the Muon System at CMS with Graph Neural Networks | Contributed talk, APS April Meeting | 2024
• Identifying Long-Lived Particles with Graph Neural Networks in the CMS Level-1 Trigger | Contributed talk, HEP-CAT Summer School | 2022

Service

• UCSD SMASH/NSF HDR ML Hackathons: co-organizer, operations coordinator, and NRP JupyterHub technical mentor (2026); anomaly-detection mentor, judging support, and event volunteer for 170+ attendees (2025)
• Conference service: Fast ML technical-support volunteer and selected poster judge (2026); CU*iP panel moderator, volunteer, and A3D3 booth representative (2025)
• Research outreach: SDSC MAP high-school cloud-chamber project lead (2023-2024); A3D3/Cross-HDR booth lead at NSBP/NSHP (2024) and SACNAS NDiSTEM (2023)
• Community STEM outreach: physics demonstrator at Barrio Logan Science & Art Expo (2022, 2023, 2025) and STEM Girl Summer volunteer (2024, 2026)

• Institute for Accelerated AI Algorithms for Data-Driven Discovery (A3D3) | 2021-Present
• NSF Career & Equity Committee
• The National Society of Hispanic Physicists | 2021-Present