Deep Learning Engineer

Our Verification team is looking for a deep learning engineer to help build cutting-edge systems for our mission to map and monitor the planet's forests. Verification covers everything from data ingestion, storage, indexing, transformations, training, inference, graphics, and tools. Each necessary to produce high-accuracy project evaluations and monitoring. As a member of the Verification team, you will research, design, implement, and deploy deep learning models that advance the state of the art in carbon mapping.
A typical day to day includes reading deep learning code/papers, implementing described models and algorithms, adapting them to our setting, working with other engineers to integrate neural networks to run efficiently, and incrementally tracking and improving feature performance.
We're looking for engineers who find joy in the craft of building and want to make an impact. Engineers who push forward initiatives by asking great questions, cutting through ambiguity, and organizing to win. Engineers who are relentlessly detail-oriented, methodical in their approach to understanding trade-offs, place the highest emphasis on building, and building quickly.

You will:

• Develop state-of-the-art algorithms in one or all of the following areas: deep learning (convolutional neural networks), object detection/classification, multi-task learning, large-scale distributed training, multi-sensor fusion, etc.
• Train machine learning and deep learning models on a computing cluster to perform carbon mapping and anomaly detection.
• Optimize deep neural networks and the associated preprocessing/postprocessing code to run efficiently on large amounts of geospatial data.
• Help develop a research roadmap to deliver on open questions and advance the performance of best-in-class models.

We are looking for strengths with:

• The team operates in a production setting. An ideal candidate has strong software engineering practices and is very comfortable with Python programming, debugging/profiling, and version control.
• We train neural networks on a cluster in large-scale distributed settings. An ideal candidate is very comfortable in cluster environments and understands the related computer systems concepts (CPU/GPU interactions/transfers, latency/throughput bottlenecks during training of neural networks, CUDA, pipelining/multiprocessing, etc).
• We are at the cutting edge of deep learning applications. The ideal candidate has a strong understanding of the under the hood fundamentals of deep learning (layer details, backpropagation, etc).
• Ability to read and implement related academic literature and experience in applying state of the art deep learning models to remote-sensing data or a closely related area.
• Familiarity with remote-sensing data such as satellite imagery, LIDAR, and radar.

We expect you to:

• learn fast and be humble.
• own solutions end-to-end.
• take part in strategic thinking and take apart problems.
• ship high-quality software.
• communicate well and document better.
• have fun.