Department of Computer Science and Engineering

Byoung Woo Yoo, a second-year student in the Department of Computer Science and Engineering at UNIST, has been honored with the grand prize at the 5th Future Finance A.I. Challenge, a highly regarded competition organized by KB Kookmin Bank. This esteemed event aims to recognize and promote advancements in artificial intelligence (AI) technologies within the financial sector. Teaming up with Joon Pyo Hong from the Korea Aerospace University (KAU), Byoung Woo Yoo formed a team named Square's Dream. Their groundbreaking solution, called 'QuantChat,' addresses the limitations of large-scale language models (LLMs) like ChatGPT, with a specific focus on enhancing quantitative analysis capabilities for corporate information disclosure. Byoung Woo Yoo (Department of Computer Science and Engineering, UNIST) has been honored with the grand prize at the 5th Future Finance A.I. Challenge. l Image Credit: Byoung Woo Yoo QuantChat is a model that complements the quantitative analysis limitations identified in LLM-based AI analysts, such as ChatGPT. It autonomously determines whether quantitative analysis is necessary to provide accurate answers to user queries. Leveraging tools like a code executor for analysis, this innovative model excels at addressing complex questions related to corporate disclosure data, including average annual growth rates and year-on-year changes. During the competition, the team presented a demo video showcasing the actual functionality of QuantChat as a chatbot web service. Byoung Woo Yoo expressed gratitude for this remarkable achievement, acknowledging the valuable experience gained from working with various language models during his research internship at the software testing and analysis laboratory. He also emphasized his dedication to future research endeavors aimed at enhancing software reliability and quality. Led by Professor Mijung Kim from the Department of Computer Science and Engineering at UNIST, the software testing and analysis laboratory specializes in developing verification technology for automatically detecting software errors. Additionally, the team conducts research on generating test codes using LLMs, such as ChatGPT. The competition finals were held on August 31, 2023, at the Lee-Sam-Bong Hall of Ewha Womans University. Out of the ten participating teams, only six teams were recognized for their achievements. Byoung Woo Yoo's team was among those awarded the Grand Prize by the Financial Supervisory Service. Additionally, other awards such as Best, Excellent, and Special awards were also presented to deserving teams.

Jaewon Kwak, a recent graduate from the Department of Computer Science and Engineering at UNIST is about to begin his new career as a software development engineer at Amazon. Mr. Kwak will be working within Amazon's Supply Chain Optimization Technology (SCOT) team based in Vancouver, British Columbia, Canada, starting at the end of September 2021. Mr. Kim earned both his Bachelor's and Master's Degrees in Computer Science from UNIST. He obtained his master's degree under the guidance of Professor Young-ri Choi (Department of Computer Science and Engineering, UNIST) in 2017. His research interests lie primarily in the fields of distributed computing, high-performance computing (HPC), and cloud. Throughout his school years at UNIST, Mr. Kim also participated in numerous supercomputing competitions to gain some in-depth, hands-on experiences on HPC. Above all, distributed computing is an area of increasing importance for the efficient execution of large-scale computing, such as machine learning. Indeed, there is a growing need for large-scale computing in workplaces, like Amazon where active large-scale computing, using big data takes place. According to Mr. Kim, participating in extracurricular activities, such as joining a student-led hacking club (HeXa, Hacker's eXciting Academy) and laboratory experiences, helped him immensely to gain the skills he needs to succeed in his science career.

Woongki Baek, Professor in the Department of Computer Science and Engineering at UNIST has been selected as a recipient of Most Influential PLDI Paper Award at the 41st ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI 2020), which took place in London, UK, Wednesday, 17 June through Friday, 19 June 2020. The award was presented in recognition of Professor Baek's research paper, entitled "Green: A Framework for Supporting Energy-Conscious Programming using Controlled Approximation," which was published in PLDI 2010. The award-winning paper by Professor Woongki Baek deals with the development of programming framework in support of approximate computing. Meanwhile, the PLDI (ACM SIGPLAN Conference on Programming Language Design and Implementation) is the top conference in the field of programming languages and compilers. In recognition of his hard work, both academically and socioeconomically, Professor Baek was given the Most Influential PLDI Paper Award at PLDI 2020. l Image Credit: Kyoungchae Kim "In general, computers are thought to carry out calculations and information processing with 100% accuracy. However, in various computing fields, such as image processing, machine learning, and web search, there is a high probability of obtaining the desired result even if there is loss-of-accuracy issue," says Professor Baek. "Approximate computing technique can be applied to these fields to achieve a desired result, while reducing the cost and energy consumption required to process massive data sets." Most Influential PLDI Paper Award is presented annually to the author(s) of a paper presented at the PLDI held 10 years prior to the award year. These papers are, then, judged by their influence over the past decade. The selection committee for the Most Influential PLDI Paper Award cited the paper as follows. "The PLDI paper “Green: A Framework for Supporting Energy-Conscious Programming using Controlled Approximation” by Woongki Baek and Trishul Chilimbi demonstrated conclusively two important things: (1) that frameworks supporting approximations that trade energy/performance costs for quality of service could be generalized and automated, and (2) applying this approach in data center search workloads could result in significant energy reduction and cost, saving millions of dollars in practice. Green proved that systematic approaches to approximate computing could have enormous commercial impact and predated and informed many subsequent machine learning optimizations, such as using reduced precision arithmetic in deep learning models. Green also inspired subsequent efforts, to more deeply encode approximation semantics into the underlying programming language design."