Hello, I'm
Building distributed systems, real-time data pipelines, and AI-powered tools in production — from edge devices to the cloud.
Hi! I'm Yuan — a Software Engineer at Corteva Agriscience, where I take full ownership of production systems from architecture to deployment. My work spans edge computing on NVIDIA Jetson, event-driven microservices, and cloud infrastructure across AWS and on-prem Kubernetes — often as the sole engineer driving projects end-to-end. I hold a Master's in ECE (ML & Data Science) from USC. I enjoy working at the intersection of backend engineering, edge computing, and AI.
Migrated 10+ sensor services from a Factory-pattern monolith to containerized microservices with standardized gRPC contracts on NVIDIA Jetson edge devices. Built a K8s Sensor Manager for ephemeral pod orchestration and replaced Redis pub/sub with NATS-based telemetry to secure edge communication.
Sole engineer owning end-to-end delivery of a pipeline processing 100+ LiDAR/multispectral flight jobs per week across 11+ containerized microservices. Shipped a PyInstaller desktop app with Azure B2C OAuth2 to 100+ field operators worldwide. Managed 4-environment infrastructure (DEV/QA/PROD/COM) on EKS with Terraform.
Automated geospatial ingestion pipeline handling 200 drone survey flights/day (20 GB each), replacing a fully manual process. Architected a Databricks pipeline to ingest orthomosaic GeoTIFF imagery via GraphQL with concurrent downloading and ArcGIS FeatureServer sync enabling downstream ML inference.
Built an MCP (Model Context Protocol) server in FastAPI that auto-injects project-specific rules and prior session history into Claude Code CLI and GitHub Copilot, so every session starts with full context. Designed an append-only audit log memory layer with cross-project semantic rules that stay in sync with the local repo.
Built a full-stack internal collaboration platform with Flask/React microservices, PostgreSQL on AWS RDS, and async email notification system via RabbitMQ. Containerized and deployed services with Docker and AWS ECS with JWT-based authentication.
Built a RAG pipeline for long-context QA with document chunking and cosine-similarity retrieval. Improved multi-choice QA accuracy by 20% on a long-context benchmark by redesigning transformer architecture with parallel encoding branches. Published at EMNLP 2023 (co-2nd author) on Vision-Language Navigation data augmentation.
GitHubI'm open to new opportunities, collaborations, and interesting engineering conversations. Feel free to reach out!
Grab a copy of my latest resume
Download Resume