Company
4 min read

Roboto AI Raises $4.8M to Build Copilot For Robotics

We're making it 10x easier to search sensor data and logs from robots and other devices!
Written by
Benji Barash
Published
April 5, 2023

Using AI to accelerate robotics development and debugging

Today, we're excited to announce that Roboto AI has raised $4.8M in seed funding led by Unusual Ventures alongside the Allen Institute for Artificial Intelligence (AI2) and FUSE Ventures. With this funding, our team is building a world-class data platform designed for modern robotics companies. We're also excited to share that we're partnering with leading researchers and professors at the ETH Zürich AI Center.

Autonomy is already here – it’s just not evenly distributed

In the coming decade, robots are going to revolutionize society. Recent advances in artificial intelligence, sensing, and embedded platforms have made this all but certain. These technologies are unlocking a new wave of automation, enabling robots to be deployed beyond controlled environments and into the real world. Every industry, including healthcare, transportation, agriculture, and aerospace, is starting to be transformed by robots. You might not have noticed yet because robots don't always look humanoid; they come in all sorts of shapes and sizes – from drones and self-driving cars to surgical arms and vacuum cleaners.

Robots of all shapes and sizes are being created in many different industries!


Every industrial transformation offers great potential for progress but also poses new challenges. It will be up to all of us: scientists, engineers, policymakers, and civic society, to develop a new framework to manage the risks and distribute the gains of these technologies. At Roboto, we believe that robots and intelligent systems should serve society and make our lives better. We want to live in a world where robots perform dangerous and repetitive tasks so that we can spend more time on creative and scientific pursuits.

Robots will also be crucial to solving worldwide labor shortages, especially post-Covid-19. Last year, the U.S. construction industry alone faced a shortfall of 650,000 roles, according to Associated Builders and Contractors. Similarly, the U.S. manufacturing industry is expected to have 2.1 million unfilled jobs by 2030. Quite simply, we're running out of skilled humans who are willing to perform these jobs, and rightly so!

Death by a thousand cuts

We've had an amazing front-row seat to the incredible potential of robotics, having spent our careers working at Amazon on projects such as drone delivery. However, we also learned that, despite all the recent progress, building safe and reliable autonomous systems is still surprisingly hard and very expensive. Many robots fail to graduate from prototypes in a lab to products in the real world.

Today, robotics companies are unfairly disadvantaged because they have to build almost everything in-house, from custom hardware to data infrastructure. Imagine you're building a new product, yet to analyze your data, you first need to hire a team to create your own version of Splunk, Tableau, or Datadog - it would be a terrible use of your resources. Unfortunately, this is the reality for most robotics companies. Standard platforms don't support robotics data, so custom infrastructure has to be developed in-house alongside limited open-source tools.

In minutes, a single robot can record terabytes of multi-modal data from cameras, lidars, and other sensors. This data gets stored in complex formats that are hard to work with and require engineering support to decode and analyze. This is time-consuming and quickly becomes a pain point for everyone working with robots. As companies collect giant volumes of data, robotics engineers have to spend their time wrangling it instead of doing robotics work.

At Amazon Robotics, we used to spend entire days writing scripts to filter and transform sensor data so that we could debug system failures, evaluate performance and create new algorithms. As we scaled up our robotic fleets, the collected data became overwhelming, and we ran into a long tail of edge cases that felt like perpetual bad luck. What frustrated us the most was how long it took to get the data we needed.

We're bringing AI-powered data tools to the robotics industry

Robotics is hard, but it shouldn't be this hard. We want to catalyze progress in the robotics industry and see robots get to production faster by getting AI to do the dirty work for us! We’re building Roboto to be an out-of-the-box solution to the data challenges engineers face daily and we're excited to have the support of an incredible group of advisors and investors to make it happen. We're building new AI-powered tools for anyone working with sensor or log data; think copilot for robotics.

Want to try it out? We’ve released a free sandbox where you can try an early version of Roboto yourself. Our design partners have already been using a version of this platform to accelerate their development. 

In the sandbox, you'll find data from nuScenes which is a large-scale, autonomous driving dataset used by engineers and researchers. We've made it easier than ever to search and explore. You can perform searches across modalities using natural language, e.g., "Find me all drives where the vehicle speed was greater than 35 km/h, and a person was detected on the right-hand side of an image".

Searching sensor data from robots using natural language in the sandbox. Sample data © nuScenes.


You can also perform searches on graphical time-series signals, something we're very excited about. Stay tuned for more on this in a subsequent post, as well as product updates and open-source initiatives that we're kicking off.

Searching for patterns in time-series data using signal search in the sandbox. Sample data © nuScenes.


Meanwhile, if you're building a robot or sensor-based product, we'd love to talk and show you more. Drop us an email at info@roboto.ai.

Or, if you're looking for your next adventure, get in touch - we're hiring!

Benji & Yves