Software QA Engineer - Data Analysis and Test Operation
Runs and analyzes software QA for FuriosaAI's NPU products in Seoul: planning test scopes, operating tests across the SDK, firmware, driver, and runtime stack, and interpreting performance, power, and stability data to find defects and report release quality. Core stack: Linux, Python/Shell scripting, and CSV/dashboard-based data analysis.
- Compensation
- Not disclosed
- Location
- Seoul, South Korea
- Work setup
- Not specified
- Seniority
- Not specified
- Employment
- Not specified
- Experience
- Not specified
- Category
- Qa
- Posted
Skills
Must have
- Software QA or test design/execution experience; can turn requirements into verifiable scenarios and pass/fail criteria
- Can define test scope and priorities considering normal/exception/boundary conditions and change impact
- Basic Linux commands and provided test tools to operate evaluations and inspect logs, processes, and device state
- Use tables, CSV, and existing dashboards to compare results and interpret distribution and variance of measurements
- Can read and lightly modify/apply existing Python or Shell scripts for test conditions and data cleanup
- Basic understanding of NPU boards, server environments, and the relationship between FW, Driver, and SDK, reflected in test conditions and result interpretation
- Can separate observed facts from cause hypotheses and clearly document reproduction conditions, expected vs. actual results, and analysis evidence
Nice to have
- QA experience with AI accelerators (NPU/GPU), SDKs, or hardware/driver-integrated software
- Experience with LLM model serving and performance benchmarking tools such as vLLM
- Experience analyzing per-version performance regressions, per-board variance, power efficiency, or long-run stability
- Basic SQL queries, statistical comparison, or use of tools like Grafana, Superset, or Metabase
- Experience analyzing how PCIe, NUMA, temperature, clocks, and throttling affect performance
- Experience improving validation scope based on customer use cases or operating release verification
Full description
About FuriosaAI
FuriosaAI builds high-performance, high-efficiency AI compute for the Inference Era. Founded in 2017 by veteran semiconductor and AI algorithm engineers, Furiosa operates globally with offices in Korea and Silicon Valley, along with a compiler-focused R&D lab in Lisbon.
Our vision is to make AI computing sustainable, enabling access to powerful AI for everyone on Earth. We solve the AI hardware energy and operational cost crisis at the architectural level, rather than through brute force, building the world's first truly AI-native compute platform to unlock the full potential of artificial intelligence for every enterprise.
About the Role
Solution QA는 FuriosaAI의 NPU 제품과 이를 구동하는 소프트웨어가 개발 환경, 생산 현장과 고객 시스템에서 요구하는 품질을 충족하는지 검증하는 팀입니다. NPU 보드와 시스템 하드웨어부터 Firmware, Driver, Runtime, SDK와 모델 추론 솔루션까지 검증 범위를 연결하고, 기능·성능·전력·안정성·호환성과 제조 신뢰성을 함께 살펴봅니다.
NPU 제품의 소프트웨어 평가를 운영하고, 평가 결과를 해석하여 불량과 품질 저하를 찾아내는 직무입니다. SDK와 관련 FW·Driver·Runtime, 모델 서빙 환경이 요구사항대로 동작하는지 확인하고, 제품 버전과 보드·서버 구성에 따른 결과 차이를 분석합니다. 평가를 수행할 수 있는 도구 활용 능력과 데이터에 근거한 판단 역량을 바탕으로 제품의 품질 상태를 설명합니다.
Key Responsibilities평가 계획과 시험 조건 관리
- 제품 요구사항, 개발자 가이드, 릴리스 변경 사항과 기존 결함을 검토하여 평가 범위와 우선순위를 정합니다. 정상·예외·경계 조건, 호환성과 회귀 항목을 구분하고 시험 목적, 사전 조건, 수행 절차, 기대 결과와 판정 기준을 정리합니다. 테스트 프로그램의 추가·수정이 필요하면 SW Engineer에게 입력 조건과 필요한 결과 데이터, 재현 시나리오를 전달합니다.
평가 운영과 진행 상태 관리
- 보드·서버·OS·Kernel 및 소프트웨어 버전 등 실행 조건을 확인하고, 제공된 프레임워크와 도구를 사용해 시험을 수행합니다. 진행 중 발생한 중단, 로그 누락, 환경 설정 오류를 확인하고, 유효한 시험 결과와 재실행이 필요한 결과를 구분합니다. 평가 일정과 커버리지, 진행률과 미검증 항목을 관리하며 조건을 통제한 재시험을 수행합니다.
성능과 전력 데이터를 통한 이상 분석
- 기능 결과, 벤치마크 성능, 전력, 온도와 시스템 로그를 함께 해석합니다. 동일 조건의 반복 측정과 이전 버전 결과를 비교하여 정상적인 변동, 성능 회귀, 서버별 차이와 보드별 이상 편차를 구분합니다. 이미 제공된 대시보드와 분석 도구를 활용하고, 필요한 경우 CSV 정리·간단한 질의나 스크립트로 분석을 보완합니다. 결과를 단순 비교하는 데 그치지 않고 실행 조건 차이가 해석에 미치는 영향을 확인합니다.
불량 재현과 개발 조직 협업
- 이상 현상의 발생 조건과 재현율을 확인하고, 제품 자체의 결함인지 서버 환경 또는 시험 프로그램의 문제인지 가설을 나누어 조사합니다. 관련 로그, 버전, 장치 정보, 기대 결과와 실제 결과를 정리해 개발 담당자에게 전달합니다. 수정 후에는 같은 조건의 재검증과 관련 회귀 시험을 수행하고 해결 여부와 잔여 영향을 확인합니다.
품질 상태 보고
- 평가 결과와 주요 불량, 영향 범위, 판정 근거와 미검증 위험을 정리하여 릴리스 검토에 필요한 정보를 제공합니다. 반복되는 분석과 운영상의 불편을 구체적인 도구 개선 요구사항으로 정리합니다. 주요 산출물은 평가 계획 및 TC 문서, 실행 조건과 결과 기록, 불량 분석 자료와 품질 검증 리포트입니다.
Minimum Qualifications
- 소프트웨어 QA 또는 테스트 설계·수행 경험이 있고, 요구사항을 검증 가능한 시나리오와 판정 기준으로 구체화할 수 있는 분
- 정상·예외·경계 조건과 변경 영향도를 고려하여 테스트 범위 및 우선순위를 정할 수 있는 분
- Linux의 기본 명령과 제공된 시험 도구를 사용해 평가를 운영하고, 로그·프로세스·장치 상태를 확인할 수 있는 분
- 표, CSV, 기존 대시보드 등 분석 도구를 활용하여 결과를 비교·정리하고, 기본적인 측정값의 분포와 편차를 해석할 수 있는 분
- 기존 Python 또는 Shell 스크립트를 읽고 실행 조건이나 데이터 정리 작업을 간단히 수정·활용할 수 있는 분