Wanyu Z.

Key Skills

Software

AWS SageMaker

Other

Top Subject Matter

Machine learning data pipelines for transit data features

Deep learning research infrastructure for training data throughput

Top Data Types

Text

Top Task Types

Data Collection

Freelancer Overview

Quantumera AI — Software Engineer Intern (AI Data Strategy). Brings 3+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include AWS SageMaker and Other. Education includes Master of Engineering; Bachelor of Science; Bachelor of Arts, University of California, Berkeley (2025). AI-training focus includes data types such as Computer Code and Programming and labeling workflows including Data Collection.

Labeling Experience

Quantumera AI — Software Engineer Intern (AI Data Strategy)

AWS SageMakerData Collection

Led ML data strategy work focused on the data pipeline used to support model iteration rather than manual annotation tasks. Owned the end-to-end ingest-to-feature workflow to determine whether downstream models had usable data. Built an AWS-based data processing stack that enabled cleaning and feature storage for ML training and consumption. • Designed and shipped ingest, batch cleaning, and feature-store infrastructure using AWS services. • Implemented schema normalization across integrated real-time public-transit API datasets. • Used Step Functions and EMR for batch cleaning and ElasticSearch as a feature store. • Deployed on ECS with auto-scaling and validated end-to-end under real traffic conditions.

2024 - 2024

Yisen Tech — Software Engineer Intern (Infrastructure for DL Research Teams)

OtherData Collection

Supported deep learning infrastructure work to unblock internal DL training by addressing storage and metadata bottlenecks that affected data access during training. Profiled system performance to locate the bottleneck specifically in centralized metadata management for small-file workloads. Developed middleware to reorganize small files into larger logical units while preserving semantics for downstream consumers. • Conducted profiling with Perf, Netperf, mdtest, and ioprof to identify performance constraints. • Traced root cause to centralized metadata management rather than raw throughput. • Shipped C middleware to wrap small files into larger logical units (POSIX truncate/wrap, NFS, loop devices). • Improved training data accessibility for internal research teams using the cluster.

2023 - 2023

Education

Work History