CV

Chang-Jae "CJ" Lee

Investigated the project's Redis usage pattern and fine-tuned lettuce and akka dispatchers to reduce async pool exhaustion warnings (since changing the architecture requires much longer time)
Implemented a new endpoints for internal users (i.e. integrators) to easily track their contents' original payload used to generate them

Implemented the all-projection handling Flink jobs to reduce resource usage and operational costs (9GB memories per EKS node running multiple pods -> 2GB)
Configured GitLab pipeline to:
- automatically generate jobs' input arguments using helmfile and its automation
- automatically generate required kafka topics and ElasticSearch indices when required

Released phase 1 on Mar. 20, 2020 (in-house log system using ELK stack)
With other 1 SDE, implemented and iterated design from initial system proposal
Techs:
- ELK via AWS ElasticSearch Service
- Lambda, CloudWatch, Kinesis, Firehose, S3, Cognito, AWS Step Function
- Yaml (CloudFormation) and Python for Lambda/StepFunction handlers, and Ruby for Pipelines configuration

Launched an initial experiment for discovering Prime Video from Amazon Prime order checkout
Techs:
- In-house Spring/JSP stack for template
- In-house A/B testing framework

Released phase 1 on Oct. 31, 2019 (NA marketplace only for now)
Managed localization string onboarding process and initial version of Offer contents storage, to reduce confusion between business and tech teams
Techs:
- Customized Java Spring + JSP
- Customized Tomcat
- Node.js on occasional scripting

Rewards for small businesses and physical stores deals: implemented the RC side features both backend and frontend, using the same tech of Prime RC

Implemented user report feedback message system, which generates 500k+ messages and 550k+ transactions per day, including Angular frontend
Developed/managed PUBG lobby’s Anti-cheat service backend, withstood peak concurrent user 3.2 million (.NET Core)
Pinpointed the exact cause of service outage when Steam auth. API is down, which was there about 1 year
Pinpointed the exact cause of memory leak in real-time game monitoring tool
Collaborated with Valve corporation’s engineers for Steam Trust Value System
(STV is already applied in CS:GO and Valve Games, not sure when PUBG or other games will be using)
Techs: .NET Core (backend), Angular (frontend), Node.js and Python 3.x (occasional scripting)
Infras: AWS

Designed/implemented server + agent
which sorts suspects by number of reports,
and streams replay on Unreal client
Techs: Vue.js on Webpack + Node.js runs on AWS EKS k8s, and a C# WinForm app.

Designed/implemented batch operation system(i.e. same actions to ~150K accounts per click, this feature now handles batch jobs targeted to ~30M users)
Participated in frontend migration: from Razor Webpage to Vue + TypeScript

Implemented ADL-to-C++ transpiler(Java, ANTLRv4),
and boilerplate for HDF5 image file processing used with transpiler's output(C++14)

Designed/implemented a subsystem which takes 2 or 6 1Gbps channels as an input and outputs a raw image from satellite or camera-in-test using C# .NET Framework 4.0
Designed/implemented a UI for camera testing and verification with moving window display which shows the output raw image in real time
Maintained/Deployed DIS for satellites already in service (KompSat 3a and KompSat 5)

Docker
Kubernetes (knowledgeable - EdX Course)
AWS
- DynamoDB, RedShift, ElastiCache Redis
- Lambda, Step Function
- Kinesis, Firehose, ElasticSearch Service
- CloudFormation, CloudWatch
Prometheus/Grafana, DataDog
GitLab CI/CD pipelines
VoltDB (JVM based RDB)
MySQL
ELK Stack (ElasticSearch, LogStash, Kibana)

Rust and WebAssembly using wasm-bindgen (knowledgeable)
Tableau (used during PUBG days, knowledgeable)
This blog is powered by Ghost runs on AWS (Lightsail -> EC2)
- Talk(Coral Project) installation is scheduled, using docker-compose