Hwaai Kang Kee

Overview: This service is used to deploy flat-file map databases to production while ensuring that the deployed map versions meet our team's threshold for data errors.

Problem:
1. Current map deployment process requires manual cross-team coordination between on-calls and QA, resulting in lost development hours.
2. Manual processes are also more error-prone.
3. Manual QA is done for every map version release, and automated map data testing should be created to support targeted weekly cadence for map updates.

What I did: I designed and developed a Java service to manage map deployments to production. This service coordinates with a dependent service to manage map deployments across multiple teams.

How service works:
1) It listens to messages in AWS SNS and SQS from a dependent service.
2) Message triggers AWS Lambda.
3) AWS Lambda triggers AWS Step Functions workflow.
4) AWS Step Functions provisions an AWS EC2 host to run extensive testing.
5) Produces an HTML file with reports for the delta between two different map versions.
6) Determines if a particular map version is suitable for production deployment.
7) Deploy the map to production servers.

Overview: The map builder is used to convert map JSON data into our built-from-scratch spatially indexed flat-file database format. These database files are then used by our navigation services to provide maps and navigation to delivery drivers both online and offline.

Problem: The current map builder process is run once a week for an hour, so the service is idle most of the time, resulting in wasted resources.

What I did: I re-architected the map building service from a long-running host architecture into a Python and Java-based event-driven architecture.

How it works:
1) AWS Lambda listens to AWS SNS and SQS for the location of a new version of map JSON data from a dependent service.
2) AWS Lambda provisions an EC2 instance to start the conversion of map JSON data to spatially indexed flat-file database files.
3) Database files will be stored in S3 after the build is finished.
4) The EC2 instance is automatically terminated after the process ends.

A Python AWS Lambda function that parses production service logs stored in Elasticsearch to identify data errors and deduplicate errors originating from the same coordinate and publish them as a CSV file for manual correction by the data team.

I designed and implemented this project as a software engineer.

A Java-based REST API microservice that aggregates data artifacts produced from multiple teams and serves compatible versions of data artifacts to the mobile client to provide offline map view, routing, and navigation to delivery drivers.

I was the back-end developer and implemented parts of the system.

A back-end Node.js service written in TypeScript that provides the Forex rate after giving two currency pairs as the request. This service uses Fixer API as an oracle and provides an in-memory caching layer to reduce Fixer API calls. It accepts both REST API requests and GraphQL requests.

A BEP-20 cryptocurrency token for education purposes. Deployed on Binance Smart Chain (BSC) and provided liquidity on PancakeSwap, making it tradeable on all decentralized finance swaps.