Software Engineer

Seattle, WA · LinkedIn · Portfolio

Skills

Server-side: Java, GCP, AWS, Docker, Node.js, MongoDB, MySQL, Memcached, Redis
Client-side: C++, JavaScript, C#, Java, Python, Shell Script, WPF, Gradle, PHP

Nov 2025 – Present

Executing migration from BrainCloud (JS) to Nakama (Go), reducing backend costs by ~75%
Optimized startup time 73% (7.8s → 2.1s) via async audio loading and deferred SDK init
Drove adoption of AI-assisted dev workflows (Claude Code) across frontend and backend
Automated manual theming process using configurable data and override rules for UI assets
Modernized a core library, adding Apple Silicon and ARM64 support and migrating GNU to GCC
Supported infra migration through cross-team testing, client/server diagnostics, error analysis

Sep 2021 – Oct 2025

Developed cross-platform features (C++, JavaScript) for Android and iOS (+ Web for testing)
Reduced Firebase Realtime Database usage with a 22.8% cost reduction
Enhanced Android build system (Gradle, Shell Script), achieving 5x performance improvement
Improved RESTful APIs (Java) with concurrency and MySQL for new social features
Implemented an auto event scheduler on the server side, replacing a manual process
Managed server deployments using Git and GCP’s Cloud Build, monitoring server health
Optimized content generation (Shell Script, PHP) by streamlining/removing unnecessary steps
Maintained 1st and 3rd-party SDKs for Android (Java, C++, JNI), supporting iOS (Objective-C)
Implemented a level design tool using WPF, significantly reducing manual design time
Prototyped the frontend/backend of a web store using React, GraphQL, Adobe Commerce

Nov 2015 – Dec 2016

Developed a compatibility pipeline between Java and C++
Created a multi-language support system in C++
Integrated external social SDKs and extended features to enhance social content
Adapted a Japanese project for Southeast Asia, collaborating with external teams

Kafka, Cassandra, S3, GraphQL, Gemini

Built event-sourced + CQRS platform: API publishes to Kafka, consumers project to Cassandra
Layered effectively-once over at-least-once via LWT dedup, retry-with-backoff, and DLQ
Isolated failure domains with two independent Kafka consumer groups: projection + enrichment
Instrumented with pino structured logs and Prometheus end-to-end + consumer-lag metrics
Built similar-incident search via Gemini embeddings in Cassandra VECTOR with SAI ANN index

Kotlin, TensorFlow, Python, AWS, Docker

Trained TensorFlow Recommenders on MovieLens 100K, served via Docker on AWS EC2
Built ratings DB → model pipeline with hourly automated retraining and cold-start fallback
Built Kotlin Android client: swipe-to-rate UI, REST (Fuel), dynamic poster loading (Bing + Picasso)

C++, AWS EC2, Node.js, MongoDB, Redis

Split auth issuer (Node) from validator (C++) — no network callback per connect
Chose dirty-flag delta sync over full-state snapshots; cut state-update bandwidth ~75%
Built reliable transport over UDP to escape TCP head-of-line blocking on the state-sync path

DigiPen Institute of Technology — B.S. Computer Science (Real-Time Interactive Simulation), Apr 2021