Agent software engineer focusing on solving agent nondeterminism through deterministic guardrails — state constraints, file-based coordination, and approval gates. Studying frontier agent and context engineering methods and the future of agent-to-agent communciation.
Building Data Engineering Agent for IBM watsonx — leveraging latest techniques on LangChain, MCP, tool-calling, and context engineering.
Accomplishments: Top 15% performance rating. Recognized for rapid two-quarter delivery of new Python SDK for IBM Cloud Data Platform. Now in public beta: https://community.ibm.com/community/user/blogs/john-wen/2025/09/30/announcing-the-watsonxdata-integration-unified-pyt
Designed and implemented a new scalable Pub/Sub and Apache Kafka queue-based hotel search system on Google Cloud, enabling the processing of 32 million search events per hour with zero dropped events and 95% cache accuracy during peak traffic.
•
Deployed autoscaling Kubernetes pods and optimized performance by building Splunk and New Relic analysis dashboards, ensuring nearly 100% production uptime.
•
Developed Maven libraries and Java Spring back-end services, improving code readability and maintainability through object-oriented design principles and JUnit testing.
Parallelized legacy SQL Server algorithms by migrating key databases to Amazon Redshift and implementing queries optimized for columnar data storage, decreasing insurance loss calculation time by 90%.
•
Developed new C# application workflows to integrate on-prem computing with AWS cloud infrastructure, advancing the company’s cloud migration strategy for big data workloads.
Designed and animated an interactive Next.js web app with Framer Motion, increasing user engagement and supporting sales teams in attracting and converting customers.