Software Engineer, Computer Vision
Job description
Monumental is automating on-site construction with robotics and software. Our mission is to make construction software-defined and autonomous - a future where beautiful, bespoke buildings go up in a single day, with minimal labour.
We're a growing team of engineers, operators and technicians from companies like Palantir, Qualcomm, Dyson, Shopify, Tesla, Meta and Google DeepMind. It's still the early days of building the company, and this is a chance to make an outsized impact on a problem that matters. From day one, you'll be working with an experienced founding team, backed by a $32M Series B led by Khosla Ventures and Plural.
Monumental is based in the city centre of Amsterdam, in a beautiful facility with an fully equipped workshop and robot testing floor. We build the whole stack in-house - hardware and software, we're explicitly not a hybrid or remote-first company - you can't build robots or walls from home, so we're onsite five days a week.
Read more on why we believe you should spend your time working on construction, and learn about our vision https://www.monumental.co/our-vision/and company https://www.monumental.co/about/. You might enjoy Based in Europe's film on how we're fixing Europe's housing shortage https://www.youtube.com/watch?v=2-Q2F0Fcx4A, or this video by Andreas Klinger of what our office looks like https://www.youtube.com/watch?v=DOo8n19XEns.
About the role
We're looking for engineers with deep knowledge of software and modern 3D computer vision. We're building an operating system to make on-site construction possible with robotics. Our software stack allows us to do everything from 3D reconstruction of the construction site (through photogrammetry), design of custom wall structures, training and deploying vision algorithms, supply chain, path planning, and robot controls and kinematics.
- We make 3D reconstructions of construction sites that let us match the architect's (ideal) drawings to the real world. These reconstructions are made in challenging conditions - low or harsh light, limited space - by operators who aren't computer vision experts, and we need a fast turnaround so our robots can get building immediately.
- Our robots place bricks and mortar with sub-millimetre accuracy in a global reference frame. These are higher accuracy requirements than autonomous driving, which means we ask more of our sensors - an off-the-shelf RealSense camera doesn't cut it.
- Our robots aren't always on stable ground. Sometimes we build a canal wall from a boat, or work off a wobbly scaffold. That means compensating live, so we can still build a beautiful building.
You can read about how we control robots with TypeScript in this post by our CTO, Sebastiaan. https://buildmonumental.substack.com/p/plans-as-data If you're new to robotics, you might enjoy reading Bouke's blog post on his experience joining Monumental as a software engineer https://bou.ke/blog/robotics/.
What you'll do
- Developing algorithms for localisation, mapping, calibration, and state estimation using state-of-the-art techniques, allowing us to drive our autonomous vehicles in the ever-changing environment of a construction site.
- Building the continuous localisation system for our robots, combining camera, depth, and IMU data to keep track of where the robot is at all times.
- Designing and building vision systems that understand the construction site: spec the right cameras and sensors, train models that judge the quality of bricks and mortar, or recognise construction objects.
- Implementing robust, production-level code that brings your algorithms to life. You own your projects from start to finish and don't approach problems like a researcher who hands a Jupyter notebook to a separate software engineering team.
- Measuring the performance of our sensing system on a continuous basis, characterising the sources of error and performing root cause analysis - even when the robot is out on a construction site and you don't have direct physical access to it.
What we're looking for
- Industry experience building and deploying 3D vision systems into production.
- Deep understanding of, and hands-on experience with, multiple hardware sensor systems
- LiDARs, IMUs, and (depth) cameras.
- Experience developing, implementing, and testing mapping, localisation and state estimation algorithms such as (VI-)SLAM, VIO, ICP, and Kalman filters.
- Ideally, some experience with real-time, deep-learning-based computer vision - e.g. multi-object tracking systems.
- Proven software engineering experience in C, C++, Python, Rust, or comparable languages - beyond doing things in a Jupyter notebook.
- Software engineering best practices: git for version control, unit testing, automated deployment.
- A strong bias for action and output. We're not a research organisation - we ship frequently and iterate fast.
- A strong sen