Staff Software Engineer - Machine Learning

Hivemapper•San Francisco, CA

12d

About The Position

Hivemapper is a decentralized global mapping data network built by 10s of thousands of people. At any given time, many 1000s of dashcams are collecting high-quality sensor data somewhere in the world and performing various computations at the edge, including state of the art computer vision tasks. 1000s of other contributors participate as AI Trainers, annotating and confirming the output of ML models. 2024 will see us 10x the number of contributors to the network. High-resolution imagery, GNSS, IMU, and other data, are then promptly uploaded for further processing and analysis (sensor fusion, 3D reconstruction, ML, human-in-the-loop annotation, etc.). Today, different data API products allow transparent and programmatic access for anyone to consume high-resolution imagery and sensor data, precisely extracted Map Features (speed signs, turn restrictions, highway signs, etc.). New data products for traffic, construction detection and more will become available later this year. Enterprise tech, mapping companies, robotaxis, rideshares, and real estate analysts represent just a few of the customers consuming data today. Our work is fast-paced, collaborative, and data-driven. We focus on solving hard problems, and we invest in our rapidly growing team and the tech to do so. Computer Vision @ Hivemapper We work on a variety of computer vision problems at the edge as well as in large compute clusters. These span object detection and stereo depth mapping to feature detection and matching, 3D reconstruction, image classification, OCR, and other general sensor fusion techniques. We solve the problem of distilling large volumes of standardized signals into highly accurate and precise insights.

Requirements

Demonstrated expertise in building ML solutions, including training and deploying models, as well as integrating them into production software systems
Hands-on experience with Image Processing and Computer Vision: Object Detection, Classification, Tracking, Localization, 3D Reconstruction, Vector embeddings, etc.
Hands-on experience with general ML and Data Mining: Clustering, Predictions, Unsupervised Methods, Ensemble Methods, Graph Optimization, etc.
Hands on experience with 3D reconstruction pipelines ( either monocular or stereo)
Strong programming and applied math skills (linear algebra, statistics, multivariate optimization)
Strong software engineering fundamentals

Nice To Haves

PhD in Computer Vision or related field
Knowledge of distributed compute systems like Hadoop/Spark
Experience with a variety of different ML frameworks ( PyTorch, TensorFlow, OpenVINO, ONNX, etc.)

Responsibilities

Help shape the CV strategy touching the full mapping stack, all the way from hardware to data insights
Balance the state of the art and bleeding edge with practicality; produce production-grade ML solutions trained on a huge corpus of standardized data that are efficient w.r.t cost and performance
Integrate ML solutions with our production systems; at the edge and in large offline clusters