Xinmei (Suzy) Xu

cv

Last updated October 2026.

Education

2025 – 2028
University of Southern California
B.S. in Computer Science · GPA 3.98 / 4.00 · Los Angeles, CA · expected Dec 2028
  • Coursework: Machine Learning, Algorithms, Data Structures, Operating Systems, Software Development, Internetworking

Research Experience

2024 – now
Undergraduate Researcher, Cooray Group — NASA Euclid Mission
University of California, Irvine · Advisor: Prof. Asantha Cooray
Scientific ML & Computer Vision
  • NASA-funded research for the ESA/NASA Euclid space telescope: deep CNNs for AI-driven strong gravitational lens identification in large-scale astronomical imaging.
  • Designed and built AgileLens, an end-to-end CV pipeline (TensorFlow, scikit-learn) over 100 GB+ of Euclid imaging: data sourcing, cleaning, preprocessing, iterative VGG16 fine-tuning, candidate ranking, and expert visual inspection.
  • Identified 130 previously unknown high-confidence strong-lens candidates (vs. ~1,000 known before Euclid); first-author paper in Astronomy & Astrophysics (2026).
  • Designed a sliding-window defect-detection algorithm for corrupted Euclid images, improving rejection accuracy by ~20% over the default catalog filtering pipeline.
Research Software & Tools for Scientists
  • Human-in-the-loop annotation and visualization tool for galaxy image labeling and large-scale candidate validation.
  • Full-stack astronomy research platform integrating pretrained CNNs, galaxy datasets, RAG and data-processing tools; CPU parallelism and CUDA over 200 GB+ of data sped up object-detection workflow by 30%.
  • Searchable galaxy-object database (PostgreSQL + React), reducing manual data-retrieval time by 60%.
  • RAG assistant over internal research documentation (ChromaDB).
  • Reusable data-processing, training and visualization modules (NumPy, Pandas, Matplotlib) deployed for 20+ researchers at NASA, Caltech/JPL and the U.S. Euclid team.
  • Collaborate across the NASA Euclid Team and Euclid Consortium on publications, conference presentations and working-group discussions.

Software Engineering Experience

Summer 2026
Software Engineering Intern, Backend
Philo Homes (AI room-design startup) · Los Angeles, CA
  • Built a TypeScript/Node.js e-commerce backend (Medusa v2) and migrated it from GCP to AWS (App Runner, RDS, S3, CloudFront), cutting cloud cost by 40% across 1,500+ SKUs.
  • Engineered a catalog sync pipeline from PostgreSQL with a custom MikroORM module for 3D/room-style metadata, cutting catalog updates from 2 hours to 3 minutes.
  • Developed admin REST APIs and a React portal with previewed CSV imports, bulk actions, and presigned direct-to-S3 uploads of 3D models served via CloudFront.
  • Set up GitHub Actions CI/CD building Docker images to ECR and App Runner with OIDC (no stored AWS keys), cutting deploy time from 30 to 5 minutes.

Selected Projects

2025 – 2026
Distributed Online Shopping System
Spring Boot · MyBatis · MySQL · Redis · AWS
  • AWS SQS absorbs burst traffic as asynchronous jobs, improving throughput and stability under high concurrency.
  • Distributed locking with Redis + Lua and the Try-Confirm-Cancel (TCC) pattern for inventory lock / confirm / rollback, fully preventing overselling during flash sales.
  • Container CI/CD on AWS ECS with Docker, AWS CDK and CodePipeline, reducing deployment time by 30%.

Publications & Presentations

  1. A&A
    Euclid Quick Data Release (Q1). XLI. AgileLens: A scalable CNN-based pipeline for strong gravitational lens identification
    X. Xu, R. Chen, T. Li, A. R. Cooray, S. Schuldt, J. A. Acevedo Barroso, D. Stern, D. Scott, M. Meneghetti, G. Despali, et al.
    Astronomy & Astrophysics, 712, A207, 2026· First author
    arXivPDFDOI
    @article{xu2026agilelens, title = {Euclid Quick Data Release (Q1). XLI. AgileLens: A scalable CNN-based pipeline for strong gravitational lens identification}, author = {Xu, X. and Chen, R. and Li, T. and Cooray, A. R. and others}, journal = {Astronomy \& Astrophysics}, volume = {712}, pages = {A207}, year = {2026}, doi = {10.1051/0004-6361/202660307}, eprint = {2604.06648}, archivePrefix = {arXiv} }
  2. ACE
    A systematic review: Deep learning-based methods for pneumonia region detection
    X. Xu
    Applied and Computational Engineering, 22, 210–217, 2023· Sole author
    @article{xu2023pneumonia, title = {A systematic review: Deep learning-based methods for pneumonia region detection}, author = {Xu, X.}, journal = {Applied and Computational Engineering}, volume = {22}, pages = {210--217}, year = {2023} }
  3. arXiv
    Euclid Quick Data Release (Q1). The Strong Lensing Discovery Engine F – Bright and low-redshift strong lenses
    L. R. Ecker, M. Fabricius, S. Seitz, R. Saglia, N. E. P. Lines, P. Holloway, T. Li, A. Verma, F. Balzer, Q. Jin, et al. (incl. X. Xu)
    arXiv:2603.28580, 2026· Euclid Consortium collaboration
    arXivPDF
    @article{ecker2026sldef, title = {Euclid Quick Data Release (Q1). The Strong Lensing Discovery Engine F -- Bright and low-redshift strong lenses}, author = {Ecker, L. R. and Fabricius, M. and Seitz, S. and others}, journal = {arXiv preprint arXiv:2603.28580}, year = {2026} }
  4. AAS
    AgileLens: A Scalable CNN-based Pipeline for Strong Gravitational Lens Identification in Euclid
    X. Xu, R. Chen, T. Li, A. Cooray
    American Astronomical Society Meeting Abstracts, Vol. 247, 319.07, 2026· Poster
    @inproceedings{xu2026aas, title = {AgileLens: A Scalable CNN-based Pipeline for Strong Gravitational Lens Identification in Euclid}, author = {Xu, X. and Chen, R. and Li, T. and Cooray, A.}, booktitle = {American Astronomical Society Meeting Abstracts}, volume = {247}, pages = {319.07}, year = {2026} }
2026
Poster — 247th Meeting of the American Astronomical Society
AAS 247
2025
Research talk — NASA Euclid U.S. Team
California Institute of Technology (Caltech)

Professional Affiliations

2025 – now
Euclid Consortium
Science Working Group: Strong Lensing
2025 – now
NASA Euclid Team

Technical Skills

Programming
Python, C/C++, TypeScript, JavaScript, Java, SQL, MATLAB, R, Lua
ML & AI
PyTorch, TensorFlow, scikit-learn, OpenCV, LangGraph, LangChain, Anthropic SDK, RAG, MCP, ChromaDB
Scientific Computing
NumPy, Pandas, Astropy, Matplotlib, CUDA, CPU parallel computing, Linux
Systems & Infra
React, Node.js, Flask, Spring Boot, PostgreSQL, MySQL, Redis, Docker, GitHub Actions, AWS (ECS, App Runner, RDS, S3, CloudFront, SQS, CDK, CodePipeline)
Languages
Chinese (native), English (professional)