NPU Software Engineer SDK
South Korean AI semiconductor and infrastructure company building inference accelerators, servers, racks, and software for production-scale AI deployment.
Funding history
Investors
About Rebellions
Rebellions develops purpose-built AI inference hardware and an accompanying software stack. Its current offerings include the Rebel100 accelerator and deployable RebelServer, RebelRack, and RebelPOD systems, designed for energy-efficient data-center inference.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will design, build, and operate SDK release pipelines, port and optimize AI models for NPU hardware, integrate open-source inference and serving frameworks, and build verification and debugging utilities. You will author developer documentation and operate its publishing pipeline.
Requirements
- Master’s degree or higher in Computer Science, Electrical Engineering, or a related field.
- Knowledge of CI/CD tools including GitHub Actions, Airflow, and Buildkite.
- Knowledge of LLM, multimodal, vision, and speech model architectures.
- Knowledge of PyTorch internals, model customization, and graph transformations.
- Python and modern C++ development skills.
- Experience integrating Hugging Face Transformers, Diffusers, and vLLM.
- Experience deploying and troubleshooting AI inference workloads in Python and Kubernetes environments.
- Experience optimizing and deploying models with CUDA, TensorRT, TensorRT-LLM, MLIR, Triton, or TVM.
- Docker and Kubernetes ML infrastructure experience.
- Knowledge of JAX or TensorFlow.
Responsibilities
- Design, build, and operate CI/CD pipelines for SDK releases.
- Port and optimize deep-learning models for NPU hardware.
- Analyze performance bottlenecks and validate numerical parity against GPU reference implementations.
- Integrate the SDK with inference and serving frameworks.
- Design model-feeding frameworks and debugging utilities.
- Author and maintain API references, tutorials, model-support matrices, and release notes.
- Operate the documentation build and publishing pipeline.
Hiring Process
Application review → online interview → onsite interview with assessment → culture-fit interview → compensation discussion → final decision.