Machine Learning Engineers (all-levels)

Company:
Location: Barangaroo NSW

*** Mention DataYoshi when applying ***

Nearmap is unique. An Australian global technology company with incredible people.


At Nearmap, we have tens of petabytes of high-quality aerial imagery (covering half a million square kilometres a year at 5-7cm resolution, and regularly captured imagery back to 2009). We produce automated 3D models of entire cities, and have recently launched our new product, Nearmap AI, which turns our visual content into semantic information to power decisions in a wide range of organisations. As a publicly listed company focussed on growth, we have both the resources to allow you to succeed in your role, and the agility (thanks to cloud-based infrastructure) to rapidly take advantage of the latest developments in the field. Nearmap is continually evolving, and you’ll need to thrive in a fast-paced environment that changes rapidly.


We are recruiting for multiple Machine Learning Engineers (at levels from Graduate to Principal) to join teams within the AI Systems group. You’ll be designing, building, and operating software systems that take petabytes of data to transform imagery to insight. Our technology stack is based on the python scientific libraries and traverses deep learning technology such as Tensorflow and Pytorch, cloud native technology such as Kubernetes, Kubeflow and Kafka, and GIS tools such as the Shapely and GeoPandas libraries. We are committed to software best practices including infrastructure as code, GitOps, CI/CD, and as much automation as makes sense.


You will work with an incredibly passionate and talented team of Data Scientists, Machine Learning Engineers and Data Engineers, and your work will interface with systems that run on Kubernetes clusters with thousands of GPUs, train large deep learning models with novel architectures, and produce sophisticated data products for a broad range of customers. The AI Systems group mixes data science, ML engineering and software engineering, and is comprised of three teams divided along responsibility lines, rather than skill sets: Data Engine, Model R&D, and Output Data and Applications.


The Data Engine team has responsibility for building the systems and managing the data for human expert labelling of our imagery and building the training and management systems that produce our deep learning models. We have a custom labelling environment built on our Map Browser product that allows our labelling team to annotate 2D, 3D and source imagery. We use algorithms in the loop to enhance the human labelling of millions of images, and automatically produce training data sets for ML that continually grow and improve in quality as we seek to build a richer machine learning derived picture of our physical world. We support the training of large, complex models on our multi-node GPU cluster. We are looking to build software and systems in place to facilitate efficient distributed and continual model training platform. This role will be critical in achieving that.


The Model R&D team designs novel deep learning architectures that leverage our rich data set, multiple dates of high-resolution imagery, multi-angle source imagery and 3D textured mesh. We are responsible for coming up with these novel architectures, designing training and evaluation methodologies, and producing the production deep learning models that are run at scale on our data. As well as multiple data sources, we ensure our models are fit for purpose within multiple geographies, and can efficiently predict on a large and growing taxonomy of labels - we use the same models to recognise objects as diverse as power poles, and large commercial and industrial buildings.


The Output Data and Applications team picks up the trained deep learning models and executes them as part of a complex DAG of post processing operations and models to create the data products that power Nearmap AI. We run at very large scale, running our algorithms to generate content that spans tens of thousands of square kilometres per day. The post processing chain is built on Kubernetes and Kubeflow and includes a range of machine learning models to perform tasks such as building footprint vectorisation and storey estimation from Nearmap’s 3D mesh, with automated quality control and human-in-the-loop review. We design, build and maintain our own infrastructure, algorithms and applications, and produce product focussed data that meets customer needs in insurance, local government, roofing and other industries.


A typical day will look like this

  • Participation in the design and scoping of greenfield projects
  • Work within a team to deliver end-to-end technical solutions — typically starting with spike sessions, onto architectural design and test creation, iteration on the solution, and ultimately deploying to production.
  • Commitment to software best practices and a strong culture of peer review.
  • May be required to supervise associates.

Experience

We are recruiting for ML Engineering roles across all three teams as we continue to evolve our systems and are open to a range of experience level from fresh Graduates, to highly skilled Principals with extensive industry experience. We also have spots for the range from software engineers interested in ML, to those with deeper experience building systems with Pytorch and Tensorflow, or other machine learning or numerical computing experience. PhDs and Masters are common in the team, but by no means necessary - we’re more interested in what you can do!


Skills

We're after exceptional candidates, who have real-world experience but is still eager to learn.


Essential:

  • Programming/Tech Environments: Ability to code in scientific python, using a Linux environment, and GIT for source control. Commitment to software engineering principles, a keen eye for clean code, and a passion for robustness and correctness.
  • Software Engineering: Working on shared codebases to produce production quality code deployed as conda packages or Docker containers.

Highly Desirable (mandatory for some roles, desirable for others):

  • Domain Knowledge – Computer Vision: Working on Machine Learning problems applied to image data.
  • Deep Learning: Applying modern artificial neural networks to solve machine learning problems using GPUs.
  • Scale: Working with large data sets, where data sets don’t fit into memory, and require multiple nodes to compute efficiently.
  • Machine Learning: A strong grasp of machine learning fundamentals (regularisation, hyperparameter optimisation, validation methods).
  • Scientific Approach: Follow the scientific method of formulating hypotheses and applying statistical tests to validate them.
  • Cloud Computing: Working on AWS or GCP using distributed virtual machines, Kubernetes, etc. is mandatory for some roles, beneficial for others.
  • GP-GPU: Using GPUs to accelerate scientific computing.
  • MLOps: Experience with operationalizing ML applications and workflows.

Personal attributes:

  • Pragmatism: While extensive knowledge of theory and best practices are highly valued, pragmatism wins over elaborate theory when it comes to shipping products that work
  • Collaboration: data science is a team sport, communicate well, share knowledge, and be open to taking on ideas from anyone in the team.
  • Attention to detail: Showing attention to detail when it counts is important... to be considered for this role Click this link and apply some basic data science skills (an astute software engineer should be able to solve it quickly with a little googling!).

Tertiary Qualifications

Formal education in a technical, data related field


Some of our benefits

Nearmap takes a holistic approach to our employees’ emotional, physical and financial wellness. Some of our current benefits include competitive pay, access to the Nearmap employee share scheme, short and long-term financial incentives, flexible working options, paid volunteer days, gym and phone rebates, and lots of development opportunities including hack-a-thons and pitch-fests.


If you can see yourself working at Nearmap and feel you have the right level of experience, we invite you to get in touch.


Watch some presentations on what we do in the AI Systems group:


https://youtu.be/WSMYfAEdAe4

https://youtu.be/ZEGdSLWdrH0

https://youtu.be/JuHBJk2uuD8

https://youtu.be/8mSSG6uICW0


Nearmap does not accept unsolicited resumes from recruitment agencies and search firms. Please do not email or send unsolicited resumes to any Nearmap employee, location or address. Nearmap is not responsible for any fees related to unsolicited resumes.

*** Mention DataYoshi when applying ***

Offers you may like...

  • Reify Health

    Senior Machine Learning Engineer
    Remote
  • AI INVESTMENTS SPÓŁKA Z OGRANICZONĄ ODPOWIEDZIALNOŚCIĄ

    Python Machine Learning Engineer
    Warszawa, mazowieckie
  • MasterCard

    Senior Machine Learning Engineer
    Dublin
  • Intel

    Machine Learning Engineer
    Bengaluru, Karnataka
  • JPMorgan Chase Bank, N.A.

    Machine Learning Engineer
    Bengaluru, Karnataka