Seoul National University

Software Engineer at Grammarly

Skills: C++ · Swift · Kotlin · Machine Learning · Rust (Programming Language) · CMake · iOS · Android · Python (Programming Language)

Grammarly

Software Engineer

● Design and implement new autocorrect, which corrects spelling and grammar, from using just text to using tab coordinate together, resulting in increased acceptance rate and decreased revert rate
● Developed swipe feature from scratch that suggests user words from gesture trajectory and context, giving Grammarly keyboard missing functionality
● Implemented inference code with neural net model on device that corrects grammar, which shipped with Samsung keyboard made success co-work between Samsung and Grammarly
● Integrated ML runtime engines including Tensorflow Lite, PyTorch, ONNX runtime to make it run on device like iOS and Android platform, combining with NLP pipelines, which is part of the autocorrect feature in Grammarly keyboard
● Made prototype of personalization that collects user input sentences and learn writing pattern to adjust suggestions

● Improved JavaCard VM to make binary size smaller to fit low end devices and performance faster, and successfully made it as base JavaCard VM platform for all smart card products
● Integrated J2ME VM on custom micro chip to run MIDlet app along secure processor on a SIM card, made possible to run applications not only on phone’s processor but also on SIM card
● Customised open source JavaScript engine to make it run on mobile phones and provided SDK for applications that can run on the JavaScript engine
● Defined GUI and IO APIs for JavaScript based applications and implemented them as library, , which is packaged together with JavaScript engine and shipped for phone’s background widget customization service

UBIVELOX

Principal Software Engineer

● Designed and implemented Just-in-Time compiler for J2ME Java VM to make applications run faster
● Ported J2ME Java VM on legacy mobile phones to make J2ME based Java applications run on phones, so that the client can do application providing service
● Implemented Ahead-of-Time compiler for J2ME Java VM to make application run faster, converting class files into native compiler compiled binary
● Implemented WIPI platform, which is another set of Java based API set, which packaged with Ahead-of-Time compiler, shipped the platform package to client for application providing service

Seungil Lee