Out of the box

Distributed inference

Whether you're working on infrastructure maintenance bots, self-driving cars, or staffless retail stores, your agents can make smarter decisions, together.

Human movement and behaviour

Easily build social robots that react to human behaviour. Our system actively tracks objects and predicts human movement, extracting pose information from individuals in real-time.

Sensor and hardware support

We support most LIDAR, stereoscopic video, and IR time-of-flight sensors; or even a combination of all three.Our system runs on the same low-power embedded hardware as Telsa's self-driving cars.

Obstacle detection and navigation

With robust object detection and SLAM capabilities built in, your agents will have no problem learning from and navigating the real-world.

Start training in 10 minutes

Step 1: Plug your sensors into the blackbox.

Connect your favourite depth sensor over USB 3.0, power via 12V DC, and connect to any network with built-in wifi, GSM or ethernet. Setup should take less than 30 seconds. The blackbox is weather-proof and consumes less than 15W of power.

Step 2: Download our sdk for C++ or Python.

The Blackbox SDK is currently restricted to existing customers and beta testers. It is due for public release in late 2017.

Step 3: Config & 'Hello World'.

Follow our 'getting started' tutorials and read the docs to set up your system and start streaming simple behavioural data.

Step 4: Integrate with ROS or learning frameworks like Caffe, Tensorflow, or Theano.

#include <black/core.h>
#include <black/vision.h>
#include <learningframework.h>

// Initialise your autonomous system, models and training routines.
YourSystem bot;

Blackbox sensors("https://youraccount.api.black.ai/");
sensors.on_update=[&sensors, &bot](shared_ptr<Black::Message> message) {
    // Train on or make decision with data


Done! You can now jump right into training your very own autonomous beer delivery robots.

Our Partners and Supporters

Questions? Ask away.

We're usually pretty busy and our team does need to sleep, so while we can't promise an instant response we'll get back to you within 24 hours.