Jaeseok Huh (Jason)

Jung-gu, Seoul, S. Korea


Projects

Google Smartphone Decimeter Challenge 2023-2024

Python | [leaderboard]

• Achieved 3.667 m of the distance error (the mean of the 50th and 95th)
• Ranked 67th out of 279, with a handle "Poisson"

March 2024 - May 2024

Leveraging Fast I/O of Unikernel in QUIC Protocol

C, C++, Python | [thesis]

• Instrumented 6 QUIC (now standardized and better known as HTTP/3) implementations for comparisons
• Ported two of them to OSv unikernel and reduced the aggregated I/O time by 1.29%-35.94%, while maintaining a simliar level of response time and request per second (RPS)

Jun 2020 – Jan 2021

Test Generator for GraphQL

Python (Graphene), Javascript, GraphQL | [code]

• Developed a module to generate query strings given a GraphQL Schema Definition Language (SDL) file
• Developed a parser to enumerate target functions and a parameter generator for them, using the Python AST module

Spring Semester, 2020

Cupeer: An Offline Socializing Service Using Mobile P2P Connectivity

Android (Kotlin) | [demo]

• Developed a peer-to-peer, real-time socializing service prototype in MVC design pattern with a decentralized matching algorithm and Google Nearby API
• Conducted an experiment on its wireless connectivity and a user survey on the idea, the prototype, and future directions
• Joint work with Adrian Steffan

Fall Semester, 2019

Starcraft II Multi-Agent RL Challenge

Python (PyTorch) | [report]

• Achieved 99.98% winning rate against an embedded AI (difficulty=4) in SC2ENV by adopting a state-of-the-art Centralized Training with Decentralized Execution (CTDE) reinforcement learning model, SMIX

Fall Semester, 2019

Web of KAIST International Students Association (KISA)

Python (Django) | [site]

• Developed a website (https://kisa.kaist.ac.kr) for the KISA with an SSO service

Mar 2019 – Feb 2020

Text-based Analysis on University Ranking

Python | [slides]

• Built an ML model to predict the ranking of USNews-top-60-colleges based on crawled articles and extracted salient linguistic features, using EMPATH, GloVe embedding, and 22K online corpus • Yielded a ranking error of 11.16

Fall Semester, 2019

Neural Networks Branch Prediction with Limited Hardware Budget

C++ | [slides] | [report]

• Achieved 0.56% of misprediction rate with 32KB storage constraint

Spring Semester, 2019
Last updated - Oct. 9th, 2024