Minsoo Cheong

• Studied recent papers on speculative decoding, keeping abreast of cutting-edge research.
• Participated in LLM team, attending seminars on LLM-related papers, enriching team knowledge and expertise.
• Analyzed and studied multiple open-source code implementations from relevant papers, bridging theory with practical application.
• Contributed to speculative decoding implementation in Llama.cpp, demonstrating ability to translate research insights into software development advancements.

Key Technologies: Kotlin, SpringBoot, Armeria, Golang, gRPC, MySQL, GCP Pub/Sub

As part of the Content Platform Task Force, responsible for the production and distribution of digitalized content in the QANDA app:

• Actively participated in development and management of extensive pipeline for digitizing and distributing various types of problem data, primarily using Kotlin and Golang. This included construction of event-driven content distribution network using GCP Pub/Sub.
• Led construction of efficient web platform for content creators, typists, and solution providers, effectively streamlining production of digital content. Specifically implemented backend server providing all APIs as gRPC endpoints using Armeria
• Played a key role in three strategically-timed product release cycles for our digital school exam data product, directly contributing to a 60-fold increase in sales by final release.
• Collaborated closely with product managers and client developers, aligning objectives and ensuring smooth execution of projects.

Additional Key Accomplishments:

• Refactored Tomcat-based authentication server into fully asynchronous Spring Webflux server, resulting in significant performance improvements. This new server handled all authentication requests for app with Monthly Active User (MAU) base of 10 million.
• Developed internal library for backend engineers within company, providing features such as authentication, testing utilities, and exception handling. This resource facilitated easier server development and enhanced functionality.

Key Technologies: Kotlin, SpringBoot, Ruby, Kafka, MongoDB, gRPC, MySQL

Microservices Transition Contributions

• Played key role in porting a large, monolithic Ruby-based backend server into microservices, facilitating independent management by Commerce, Content, and O2O teams.
• Authored library named "Ozingger" that drastically simplified and automated the process of migrating to microservices. Ozingger enabled automatic searching and resolution of dependencies among microservices with minimum requests to each other, boosting performance of data aggregation from these services.
• Constructed "aggregator" server using Ozingger and Spring Cloud Gateway, which served as gateway in front of divided services and resolved dependencies. This highly scalable server handled over 60 million daily requests (750 requests per second) for commerce APIs alone, not including content APIs, which had their own dedicated aggregator.
• Successful implementation of aggregator server smoothened transition to microservice architecture, and was later used to build Backend-for-Frontend (BFF) servers by Global team.

Commerce Service Team Contributions

• As part of Commerce Service Team, collaborated closely with Product Owners and client developers to build numerous commerce-related features which included providing APIs with both REST and gRPC endpoints
• Led development of "Product Groups" feature using Kotlin, Spring Boot, Kafka, and MongoDB. This feature allows operators to manage and perform mass operations on over 100,000 store products in bulk, such as adding badges or tracking indicator measurements in groups.
• Participated in addition of various features like scheduled delivery and spelling correction in search functionalities.
• Contributed to refactoring of store feed into microservice architecture using Kotlin and Spring Boot, resulting in robust, diverse feed that provides list of various items like Products, Deals, or Advertisements, and significantly shorter load time.