# 자기 소개 Introduction --- ### Greeting - Hi, I am Gary Noh and I am currently working at Samsung Electronics as a software engineer. Thank you for having your time for my interview today. ### About me - Let me tell you about my career. For last 4 years in my company, I ve been a full stack engineer for chip verification. And I ve also designed an architecture for machine learning pipelines in **distributed systems**. - As a proof of my work, I have published some papers on top conferences. So, I can say that my major expertise is to build and do research on **systems for machine learning and machine learning for systems**. ### My goal / Motivation - I believe my experiences and knowledges can contribute to build more robust and efficient systems for Google's product such as **Tensorflow Extended**. And you may wonder why google? - I think Google already has powerful and monolithic AI models. Then, what is the next step for AI? I think the next AI will be on-device machine learning and decentralized learning with continuous training. These features will require more efficient, more cooperative and more robust infrastructure for service. Therefore, I believe Google is the perfect place to make my idea real and put my whole efforts for the next AI. Thank you again for your time and let s get into the interview! Projects --- #### 내가 한 프로젝트 중에서 가장 힘들었던 부분과 어떻게 해결했는지? (흥미로웠던 프로젝트) 1. **In-memory machine learning model architecture** - I can say 'in-memory machine learning model architecture' is one of the most interesting projects I have involved. Our team struggled with very low performance and usability of NLP model. Including loading time from disk and job dispatching time to the designated servers, it took one minute average to get the result. To improve this problem, I proposed an architecture to build our own backend systems to keep the model in the memory and available via REST API. Moreover, I used nginx for caching data to speed up inference time. As a result, The inference time has been reduced to only 1 or 2 seconds. 2. **Resource Management with Reinforcement Learning** My interests --- - TFX, Tensorflow Extended - 어떤 구글 팀에서 일하고 싶고 왜? - **[Full Stack Software Engineer, TensorFlow Extended](https://careers.google.com/jobs/results/141714149546566342-full-stack-software-engineer-tensorflow-extended-english/?distance=50&location=Seoul,%20South%20Korea&q=Software)** ![](https://i.imgur.com/qYvCbby.png) - **[Cloud Engineer, Infrastructure, Google Cloud Professional Services](https://careers.google.com/jobs/results/141443478451036870-cloud-engineer-infrastructure-google-cloud-professional-services/?distance=50&location=Seoul,%20South%20Korea&q=Software)** Questions --- - What is your job at Google Korea? - If Tensorflow Extended, what is the major issue or problem in your team these days? Do you guys have the same concerns on monolithic AI? - What is your culture for self-developed tools? - What do you guys do when you have free time? For example, studying for themselves or extra - Any tips I can get a job from Google?