JaeGwon C.

LLM Evaluation & Data Contractor

Seongnam-si, Korea, Republic of

Experience

Apr 2025 - Present
8 months

LLM Evaluation & Data Contractor

Outlier AI; DataAnnotation Technologies; Invisible Technologies

  • Worked in domains including coding, STEM, generalist, and KR/EN bilingual (voice + text)
  • Real-time face detection & recognition
  • Authored high-fidelity system/user prompts and rubric frameworks to improve evaluation research foundation consistency
  • Performed structured comparative judgments (A/B, multi-turn) with rationales; compared and refined AI responses based on rubrics
  • Produced and reviewed KR/EN voice recordings, transcripts, and others' work
  • Identified recurring model failure modes (logical chain breaks in STEM, hallucinated code libraries, bilingual register shifts) and escalated pattern summaries leading to rubric refinement
  • Created reusable error taxonomy (hallucination / reasoning gap / spec noncompliance / tone mismatch / unsafe content leakage)
Dec 2020 - Present
5 years
Korea, Republic of

Software Engineer (Backend – Ad Server)

Kakao Corp

  • Backend developer working on an ad server
Apr 2020 - Dec 2020
9 months
Korea, Republic of

Trainee

SW Maestro (11th)

  • Implemented face detection in a web browser
  • Worked on video conferencing
Dec 2019 - Mar 2020
4 months
Korea, Republic of

Software Engineer (Internship)

Kakao Enterprise

  • API gateway development
  • Search in Hamming space
May 2019 - Oct 2019
6 months
Korea, Republic of

Student Researcher

Parallel Software Design Lab, University of Seoul

  • Real-time face detection & recognition
  • GPU computing

Summary

Results-driven bilingual (KR/EN) LLM evaluation specialist and backend engineer (4+ years at Kakao Ad Server). Strengths include high-fidelity prompt and rubric design, structured comparative judgment (A/B, multi-turn), and clean multilingual (voice + text) data generation. Earned more than USD 10,000 as a consistent contributor across coding, STEM, generalist, and bilingual evaluation streams. First author in real-time face recognition; proven algorithmic problem solving (Grand Prize, UOS). Blend of production and review latency-aware systems mindset and linguistic nuance for scalable model quality improvement.

Failure mode identification (logic chain breaks, hallucinated code libraries, register shifts)

Error taxonomy (hallucination, reasoning gap, spec noncompliance, tone mismatch, unsafe leakage)

Languages

Korean
Native
English
Advanced

Education

Oct 2015 - Jun 2021

University of Seoul

Bachelor's Degree · Electrical & Computer Engineering · Korea, Republic of · 4.04/4.50

Need a freelancer? Find your match in seconds.
Try FRATCH GPT
More actions