Resume

Education

  • B.S., Applied Mathematics, University of California, Irvine
  • B.S., Computer Science, University of California, Irvine

Selected coursework

  • AI/ML (CS 171/175/178): Supervised and unsupervised learning (178), search/planning (171), NLP (175).
  • Computer Vision (CS 116/117): learned image processing (116) and 3D reconstruction (117).
  • Optimization (Math 110A/B): linear, non-linear, convex optimization.
  • Probability theory (MATH 130A/B/C): went over discrete, continuous probability distributions and stochastic processes.
  • Linear Algebra (Math 121A/B): proof based linear algebra, covered vector spaces and inner product spaces.

Industry Experience

AI Engineering Intern — Cast & Crew

Burbank, CA • June 2026 — Present

  • Developing contract extraction pipeline using an LLM and the Unstructured API to pull structured data from unstructured, mixed-format contract documents
  • Designing the pipeline to integrate with multiple databases (SQL Server, Postgres) inherited from acquired companies, unifying contract data into Slate's consolidated platform
  • Coordinated with engineering teams across acquired companies to map undocumented database schemas, translating findings into the pipeline's data integration design

Machine Learning Engineer Intern — MyFitnessPal

Remote • June 2025 — Aug 2025

  • Shipped image-generation pipeline populating the food database, using dedicated images for high-traffic foods and taxonomy-level images for all others
  • Reduced "no match found" in production meal classifier by 4% by mining 10K+ misclassified examples from inference logs and retraining on hard negatives
  • Benchmarked Logistic Regression, DNN, CNN, and ViT classification architectures on ROC-AUC and inference latency; selected DNN on DINOv2 embeddings for 95% accuracy at sub-200ms latency
  • Built SQL + Pandas pipelines over Snowflake processing 1M+ records to construct the training and evaluation datasets for ML modeling
  • Deployed meal classifier via Streamlit on Dockerized GPU node as internal evaluation tool adopted by data science and product teams

Research Experience

Machine Learning Undergraduate Researcher — University of California, Irvine

Irvine, CA • Oct 2024 — June 2026

  • Authored custom CUDA C++ kernel for novel structured pruning with MUX rewiring on various LLaMA-3 models, achieving 5x speedup in pruning runtime over PyTorch reference implementation
  • Developing novel structured pruning method achieving 30% perplexity reduction over baseline structured pruning
  • Built config-driven experiment framework in Python orchestrating 2,000+ parallel GPU jobs on SLURM-managed HPC cluster, automating pruning, LoRA finetuning, and evaluation across model and sparsity configurations
  • Reproduced efficiency methods including Wanda, AWQ, LoRA, and logit distillation