Career Profile
I’m a ML engineer interested in deep learning.
Currently, I’m working as a skilled industry personnel.
Education
Work Experience
- Developed a central content moderation system for global services of Match Group.
- Developed an AI chatbot for dating simulation.
- Implemented deepfake system using an identity swapping engine that manipulates the face and voice of the target video. Improved pose robustness and attribute preservation by using NeRF and GAN-Control for data augmentation.
- Implemented 3d digital human using a TTM (text-to-mesh) engine that generates talking 3d human mesh from text. Improved mesh generation accuracy by optimizing the architectures of landmark detection model and 3d reconstruction model.
- Implemented super-resolution model for video restoration. Used a GPEN-based model and improved the resolution of the video up to 4x (from 256 to 1024).
- Designed and implemented an integrated research environment including a new computer vision framework based on pytorch lightning. Improved training speed up to 7 times with multi-gpu training using DDP.
- Redesigned the cash log table of Cashwalk (MAU 6 million) to an optimized NoSQL schema and migrated it from RDS to DynamoDB.
- As a result, cloud infra costs were saved by an average of 9% per month.
- Designed and implemented a survey website to evaluate the services of hospital staff, creating economic value of more than $20,000.
Extracurricular Activity
- Studied meta learning model for personalized diagnosis of cardiac abnormality using 12-lead ECG classification.
- Optimized the architecture and hyperparameters of transformer-based model and improved inference accuracy by more than 20% (from 0.62829 to 0.76013).
- Participated in the PhysioNet/CinC Challenge 2021 and achieved the highest test accuracy (70.2%) among 40 teams on the test dataset.
- Studied automatic code generation of script language for service-oriented IoT platform using fine-tuned GPT-2.
- Developed a backend server of service-oriented IoT platform (about to be commercialized).
- Applied TensorRT to the inference pipeline and improved inference latency by more than 8 times (from 7683ms to 936ms).
- Designed and implemented SNUGH, a website that helps college students plan their courses beforehand and manage graduation requirements.
- Participated in backend development, mobile app development and DevOps management.
- Designed and implemented Proudly, a mobile app developed for self-management alarm service using photo authentication.
- Participated in mobile app development.
- Designed and implemented Finger Call, an mobile application that helps people be on time by calculating late fees and sharing each other’s current location in real time.
- Participated in mobile app development.
- Taught overall android development, clean code and clean architecture as the mentor of android seminar.
- Succeeded in attracting small investments for group blind date matching and restaurant reservation service named Meeting Playlist.
Open Source
A promise-based Node.js ORM for DynamoDB inspired by Sequelize.
Certificate
Skills
Machine Learning
- Language (Python, C++)
- Numpy & Pandas
- Scikit-learn
- PyTorch & PyTorch Lightning
- TensorFlow 2.0
- Optimization (CUDA Toolkit, TensorRT)
- Visualization (TensorBoard, Wandb)
- MLOps (MLFlow)
Backend
- Language (Python, JS)
- Django & DRF
- Node & Express
- ORM (Sequelize, Dynamize)
- DBMS (MySQL, PostgreSQL, MongoDB, Redis, etc)
Infrastructure
- Docker & Docker Compose
- CI/CD (Github Action, Jenkins, CircleCI)
- AWS (EC2, EB, RDS, DynamoDB, ElastiCache, S3, CloudFront, Lambda, Code Pipeline, Route53, CloudWatch, etc)
- GCP
Android
- Language (Java, Kotlin)
- Asynchronous Programming (RxJava, Coroutine)
- Data Binding (LiveData)
- REST Client (OKHttp, Retrofit)
- ORM (Room)
- DI (Koin, Dagger)
- MVVM & Clean Architecture
Collaboration
- VCS (Git, GitHub, GitLab)
- API Docs (Swagger, Postman, Stoplight)
- Wiki (Notion, Confluence)
- Issue Tracking (Jira, Trello)
- UI/UX Design (Figma, Zeplin)
- Communicaiton (Slack, Microsoft Teams)
I’m apt at using various collaboration tools for effective communication with other developers, UI/UX designers, product managers, etc.