Staff Machine Learning Infrastructure Engineer (1525)

About the Role: The Vision & Learning team is responsible for building out the future of Matterport's core technologies, spanning the 3D reconstruction of spaces from physical data to the semantic understanding of our digital twins. Along with our ever-growing user base, the scales at which we need to apply these technologies and the massive datasets of complex real-world data upon which we're refining them are expanding rapidly. We see the challenges of these scales as opportunities, and we believe that efficient and robust infrastructure is central to surmounting them.

Our ideal candidate for this role sees this opportunity and is passionate about building out the infrastructure that can make the most of it. They'll find excitement in accelerating the capture of small apartments with mobile devices and are eager to navigate the complexity and compute resources involved in correctly reconstructing massive museums and stadiums alike. They understand that data is essential to powering the machine learning approaches to such problems, and are already planning out how we'll train upon ever more examples per second.If grappling with terabytes of data, scaling cutting-edge computer vision and machine learning technologies, and creating value for real users appeals to you, then we’d love to talk!

What you will do: Working side-by-side with the computer vision and deep learning engineers on the team and reporting directly to the VP of Software for Vision & Learning, you'll be responsible for accelerating the research and productization of Matterport's core technologies.

You may expect to: advocate for tools and technologies to accelerate research organize and build out systems to make data accessible to the team identify common threads in team activities and standardize experimental workflows, emphasizing metrics-driven research distribute neural network training across fleets of compute servers to leverage massive datasets optimize machine learning inference on mobile devices for near real-time use liaison between research and platform teams at Matterport to facilitate rapid deployment and iteration of products

Who you are: Some baseline experience and skills necessary to succeed in the role 4+ years of software engineering experience Experience building scalable backend services or data processing pipelines Solid understanding of ML/DL fundamentals and production ML workflows Ability to write performant, well-tested, production-ready Python code Additional experience and skills that will help you succeed: Experience with Docker/Kubernetes/Terraform/AWS/GCP Experience training machine learning models using PyTorch or Tensorflow Experience building deep learning pipelines including: Metrics collection, experiment tracking and hyperparameter tuning (e.g. W&B, Comet.ML, DVC.org) ML lifecycle platform utilization (e.g. SageMaker, Metaflow, MLflow) Distributed training (e.g. DeepSpeed, Horovod)Experience optimizing model inference e.g. via quantization/pruning Experience deploying trained models for inference on either: Cloud, addressing issues of compute autoscaling Mobile, addressing issues of model conversion (i.e. for iOS / Android) Experience developing Python bindings or cloud interfaces to tooling written in C++