Software Engineer, Compiler (Front-end)

FuriosaAI · Software · Seoul, South Korea
공고 번호 4005784201

PyTorch 등 외부 모델을 내부 표현으로 변환하고 그래프 최적화·kernel 언어 설계.

학사 명시 · 도전 지원

확인할 점: 컴파일러/변환 infrastructure 및 Rust/C++ 개발 경험이 현재 증거에서 확인되지 않음.

영문 이력서 PDF ↗영문 이력서 Overleaf ZIP ↓

Cover Letter PDF ↗Cover Letter Overleaf ZIP ↓

이 공고의 지원서 열기 ↗

Career Summary

필수 Career Summary 칸에 붙여 넣기.

I am an ML engineer and AI Researcher at ISPARK R&D Center responsible for TieMing's core AI features, from model training to inference APIs and deployment. I bring a multimodal research background—five publications, four as first author—and turn experimental results into working products. I implement models in PyTorch and validate how input representation affects their executable behavior.

I improved normalized database-result agreement from 26.6% to 76.6% on a 64-question internal CRM development benchmark through teacher-student training and LoRA. I also verified and documented the schema input required to reproduce the model's behavior in serving.

I expanded text-only analysis into multimodal fusion and neural architecture search, producing five papers, four as first author. I applied that experience in implementing and comparing models to company AI features.

I bring a model-development perspective to compiler front-end work: implementing multimodal architectures in PyTorch, making input assumptions explicit, and checking outputs against task requirements. These experiences are directly relevant to validating model representation and support. At FuriosaAI, I can contribute representative model cases and reference checks for model support. I would trace inputs, shapes, and operators through the front end and turn unsupported behavior or output discrepancies into focused implementation work.

Cover Letter

PDF 첨부 또는 Enter manually에 붙여 넣기.

Dear FuriosaAI Hiring Team,

I am applying for the Software Engineer, Compiler (Front-end) position (Job ID 4005784201). I bring a model-development perspective to compiler front-end work: implementing multimodal architectures in PyTorch, making input assumptions explicit, and checking outputs against task requirements. These experiences are directly relevant to validating model representation and support. Before ISPARK, I served one year as an information-systems NCO in the Republic of Korea Air Force (Dec 2021–Dec 2022), maintaining Linux servers through shell scripting and log analysis.

I improved a four-billion-parameter Text-to-SQL model from 26.6% to 76.6% normalized database-result agreement on a 64-question internal CRM development benchmark, using teacher-student training and rank-16 LoRA. The model matched 49 of 64 reference results, compared with 17 for the base model; scoring ignores row order and normalizes numbers and dates. I also verified the serving interface: missing schema input caused invented table names despite unchanged weights. I documented the required schema inputs alongside the adapter so the trained behavior could be reproduced in serving.

I authored five research papers, four as first author, by developing an initial comment-analysis project into multimodal learning research. Text alone could not represent the accompanying video and audio, so I expanded the representation and investigated neural architecture search for choosing fusion paths. This required turning a research question into model code and comparing alternatives under a defined search space. I now apply the same approach to product development: choose a concrete limitation, implement a targeted change, and measure the result.

At FuriosaAI, I can contribute representative model cases and reference checks for model support. I would trace inputs, shapes, and operators through the front end and turn unsupported behavior or output discrepancies into focused implementation work.

Sincerely,
Jeonghun Kim