Data Scientist, New Grad - Model Optimization
Core
Full-time data scientist role focused on model optimization for Quadric's custom GPNPU architecture, specifically owning quantization work for production models and building numerical accuracy testing infrastructure.
Role type
New Grad Data Scientist (Model Optimization)
Builds
Quantization workflows and numerical accuracy testing infrastructure for vision and language models running on edge and endpoint devices.
Domain
Edge computing, Neural Network Inference, Hardware-Software Co-design
Deliverable
production ML models
Required skills
Python, PyTorch, NumPy, Machine Learning foundations, CNNs, Transformers, Quantization, Fixed-point arithmetic
Preferred skills
PTQ, QAT, TF-Lite, ONNX-Runtime, TVM, MLIR Quant, Embedded systems, DSPs, GPUs
Technologies
PyTorch, TensorFlow, NumPy, Matplotlib, Plotly, ONNX-Runtime, TVM, MLIR
Responsibilities
Develop and deploy quantization workflows for vision and language models; Investigate per-layer numerical error and propose calibration strategies; Extend quantization library with new operators and algorithms; Build and maintain numerical accuracy testing infrastructure; Collaborate with graph compiler, kernel, and hardware teams.
Seniority
New Grad, entry-level IC
