Algorithm - AI System Researcher

FuriosaAI
Seoul
Full timeFunction: Research & Scientific (R&D)Education: mastersSkills: ["Analytical thinking","Problem structuring","Hypothesis-driven research","Communication","Presentation"]

Research next-generation NPU serving systems by identifying cluster-level and heterogeneous computing challenges, then exploring design spaces with analytical modeling, roofline analysis, cost modeling, and a custom simulator. Validate key hypotheses directly on your NPU, developing and testing serving algorithms such as scheduling, batching, KV cache management, and disaggregation. Turn promising ideas into POCs/prototypes in collaboration with the SW organization for productization.

Loading

Loading job details...

Preparing the role view and application actions.

FursaFursa
FuriosaAI
FuriosaAI
2 weeks ago

Algorithm - AI System Researcher

✓ Verified Job

Canonical indexed version, validated from employer's careers page.

Source: Company careers pageValidated by: Fursa AI
Last checked: 44 minutes agoStatus: Live
Reposted: similar role first listed 1 month ago

Job Summary

Research next-generation NPU serving systems by identifying cluster-level and heterogeneous computing challenges, then exploring design spaces with analytical modeling, roofline analysis, cost modeling, and a custom simulator. Validate key hypotheses directly on your NPU, developing and testing serving algorithms such as scheduling, batching, KV cache management, and disaggregation. Turn promising ideas into POCs/prototypes in collaboration with the SW organization for productization.
Location: Seoul
Employment Type: Full time
Job Function: Research & Scientific (R&D)

Key Responsibilities

  • •Identify major technical challenges for next-gen NPU serving systems (e.g., cluster-level serving and heterogeneous computing).
  • •Explore the design space using analytical modeling, roofline analysis, cost models, and a custom simulator to produce quantitative performance/cost evidence for key decisions.
  • •Conceive serving algorithms (scheduling, batching, KV cache management, disaggregation), validate them through modeling and NPU-based implementation, and mature viable ideas into POCs/prototypes.
  • •Develop research directions independently and define challenging problems suitable for product application within 6 months to 2 years.
  • •Collaborate with the SW organization to move prototypes into the productization stage based on validated research outcomes.

Key Requirements

  • •M.S. (or in-progress) in AI/ML, Computer Architecture, or a related field, or equivalent experience.
  • •Understanding of LLM inference fundamentals and performance characteristics (attention, KV cache, prefill/decode, batching, multi-chip parallelism).
  • •Ability to perform performance modeling, simulation, and data-driven analysis using Python and PyTorch.
  • •Skill in structuring ambiguous problems independently and driving a hypothesis–validation research cycle.
  • •Clear communication of analysis results and technical judgments through documents and presentations.
Experience:AI/MLLLM inferenceNPU systemsAI datacentersHigh-performance computing
Education:Master's
Skills:Analytical thinkingProblem structuringHypothesis-driven researchCommunicationPresentation
Languages:English
Tech Stack:PythonPyTorchVLLMSGLangTensorRT-LLMNPUGPUSimulatorRoofline analysisAnalytical performance modelCost modelKV cacheSchedulingBatchingDisaggregationMulti-chip parallelism

Company Brief

FuriosaAI
Designs and develops data‑center AI inference accelerators (RNGD) and a full stack hardware‑software platform to deliver energy‑efficient, high‑performance AI compute for enterprise and cloud customers.
Industry: Hardware Devices
Company Size: Medium (51 to 250 employees)
Growth: Scaleup
Valuation: USD 500M to 1B
Funding: Series C
Headquarters: Seoul, South Korea
Founded: 2017
Glassdoor
Glassdoor: 4.6
WebsiteLinkedInGlassdoor