ML Research Engineer
Core
Drive software development and model optimization tooling to streamline deployment of quantized LLMs, analyze performance bottlenecks, and prototype low-precision data formats on novel hardware.
Role type
ML Research Engineer (Model Optimization & Quantization)
Builds
Quantized LLM deployment tooling, low-precision data format prototypes, optimized model architectures
Domain
AI/ML, Large Language Models, Hardware Acceleration
Deliverable
production ML models
Required skills
Model quantization, low-precision numerics (FP8, INT8/4), GPU/NPU kernel development, Transformer/LLM architecture optimization, Python, C/C++, PyTorch, CUDA, Triton, ROCm, TensorRT, ONNX Runtime, model parallelism, Flash Attention
Preferred skills
Custom BLAS kernel implementation, mixed-precision arithmetic unit knowledge
Technologies
PyTorch, TensorFlow, TensorRT, ONNX Runtime, CUDA, Triton, ROCm
Responsibilities
Develop and optimize tooling for quantized model deployment, analyze performance bottlenecks in LLM architectures, prototype low-precision data formats, co-design model architectures for low-precision deployment, collaborate with hardware and AI software teams on system efficiency
Seniority
Mid-Senior, hands-on IC