Search
Go
AI Modelling and Simulation Engineer
Summary by AIEngineer.careers
AI Modelling and Simulation Engineer is a AI Infrastructure Engineer role (full-time). with Lumai. in OXFORD, GB. Imported listing (source: earlystagestartups.com). Apply on the employer's site (earlystagestartups.com).
Imported listing (source: earlystagestartups.com) · Apply on earlystagestartups.com
OXFORD · GB
—
Full-time
SENIOR
The sections below reproduce the third-party job description for reference. AIEngineer.careers does not write or control this text.
Imported job description
Sourced from earlystagestartups.com
Lumai is seeking a Senior AI Modelling and Simulation Engineer to build the performance and power modelling framework for its optical computing hardware. This role bridges architecture and compiler teams by predicting latency, throughput, and power for photonic AI inference workloads. You will design Python-based simulation tools that characterize operator-level workloads against hardware and RTL results to drive critical product decisions.
Responsibilities
- Design and maintain a Python-based framework for modelling hardware/software latency, concurrency, and event-driven relationships.
- Model compute engines, interconnects, memory interfaces, and accelerator blocks across device configurations.
- Characterize AI operators (convolution, attention, etc.) against hardware, RTL, or emulation data.
- Build compiler-agnostic interfaces to ingest compiled models and MLIR from internal flows.
- Develop block-level power models and validate worst-case scenarios with micro-architecture teams.
- Own CI/CD pipelines for simulation flows to ensure reliability and reproducibility.
Requirements
- Proven experience building or extending hardware/software performance modelling frameworks for AI/ML accelerators in Python.
Lumai
Oxford · GB · 28+ employees
Lumai is an Oxford University spinout developing optical-computing hardware for AI inference and data-centre deployment. Its 3D optical technology performs AI processing with light rather than conventional electronic computation, targeting substantially higher performance and lower power consumption.
