Member of Technical Staff - 3D Machine Learning

Machenta; Greater London; Hybrid; GBP 90,000 - 130,000; Full time

Benefits offered by this job

Equity

Private health insurance

Free food

Job summary

Machenta Ltd in London is seeking a Member of Technical Staff to push transformer-based models for instant robot learning, building rich 3D representations from multimodal inputs, and running careful experiments to understand scaling and performance.

You’ll turn research insights into efficient models that run on real robot hardware, collaborating closely with the team to ship improvements and measure impact on robot capability.

Qualifications

• Deep expertise in modern ML, with a strong track record of designing and training transformer-based models.
• Hands-on experience with 3D representations and 3D deep learning, such as point cloud networks or multi-view models.
• Strong understanding of 3D geometry and computer vision fundamentals, including camera models, multi-view geometry and depth estimation.
• Strong Python and PyTorch skills, and experience writing clean, reliable code.
• Experience running large-scale training, including distributed training and efficient data pipelines.
• A rigorous, experiment-driven mindset: careful baselines, ablations and clear conclusions.
• Clear communication of your ideas and findings to the team.

Responsibilities

• Push the frontier of model design for instant robot learning, with a focus on transformer-based architectures and 3D representations.
• Design and run experiments on efficient multimodal sensor fusion, combining inputs such as RGB, depth and proprioception.
• Run ablations and investigate scaling behaviour across model architectures, data and compute.
• Explore new objectives, auxiliary losses and sources of supervision that make learning more data-efficient.
• Turn research insights into models that run efficiently on robot hardware and measurably improve real-world performance.

Skills

Transformer models

3D representations

3D geometry

Python & PyTorch

Distributed training

Experiment mindset

Clear communication

Job description

About the Company

We've been building Machenta for less than a year, and so we're still right at the beginning here. We're a very close team, sharing the journey together, working in person in London, and collaborating throughout. This has been great fun! And as we bring in the next few team members, we're now thinking carefully about what roles we need, and who we should be speaking with. Everybody joining will be Machenta's expert in their part of the team, with ownership and responsibility for shaping where we go next and how we get there. We pay well, and we'll be giving you meaningful equity, private health insurance, and lots of free food.

About the Role

As a Member of Technical Staff working on 3D machine learning, you'll refine the model architectures at the heart of instant robot learning. You'll explore how transformer-based models can build rich 3D representations of the world from multiple sensor inputs, and which architectures, objectives and sources of supervision let them learn the most. You'll run careful ablations and scaling studies to understand what really drives performance, and turn those insights into better models. It’s a research-driven role at the frontier of machine learning with a practical purpose: every idea you explore is aimed at making our robots more capable. You'll work closely with everyone at Machenta to see your model improvements come to life on real hardware, and the measure of success is how much better our robots learn and perform.

In This Role You Will:

• Push the frontier of model design for instant robot learning, with a focus on transformer-based architectures and 3D representations.

• Design and run experiments on efficient multimodal sensor fusion, combining inputs such as RGB, depth and proprioception.

• Run ablations and investigate scaling behaviour across model architectures, data and compute.

• Explore new objectives, auxiliary losses and sources of supervision that make learning more data-efficient.

• Turn research insights into models that run efficiently on robot hardware and measurably improve real-world performance.

What We Hope You’ll Bring:

• Deep expertise in modern machine learning, with a strong track record of designing and training transformer-based models.

• Hands‑on experience with 3D representations and 3D deep learning, such as point cloud networks or multi‑view models.

• A strong understanding of 3D geometry and computer vision fundamentals, including camera models, multi‑view geometry and depth estimation.

• Strong Python and PyTorch skills, and experience writing clean, reliable code.

• Experience running large‑scale training, including distributed training and efficient data pipelines.

• A rigorous, experiment‑driven mindset: careful baselines, ablations and clear conclusions.

• Clear communication of your ideas and findings to the team.

Bonus Points If You Have:

• Experience applying 3D perception or representation learning to robotics, such as manipulation or policy learning.

• Experience with multimodal models, sensor fusion or vision‑language‑action / world models.

• Experience with self‑supervised learning or pretraining on large‑scale visual or 3D data.

• Experience optimising models for real‑time inference, for example through quantisation, distillation or efficient architectures.

• Publications at leading ML, vision or robotics venues.

• A PhD in a relevant field.

• Experience at an early‑stage robotics or AI company.

Joining Us

We're setting off to change how robots learn, and how people interact with them. If you're excited about joining in on the adventure with us all, and about working in a close‑knit, highly ambitious, highly passionate team in the heart of King's Cross London, we would love to hear from you!