Latest Updates and Insights
In our last few updates, we demonstrated how a robot developer can use Xpy AI to run experiments and track metrics across runs. This will help robot devs perform integrated real-world robot tests with ease. However, we were continuing to think about the problem of friction in robot software development...
Read More →In our last update, we showed how Xpy integrates with the typical robot software architecture. We demonstrated this on a raspberry pi logging data that the Xpy web-app plotted in real time. Notably, it is important for the user to be able to specify when they want data to be...
Read More →As of our last update, we have created an MLOps suite that manages data from robot experiments and allows the user to compare arbitrary metrics across experiments/runs. We demonstrated this with fake data from a laptop, but in this demonstration, we will be logging data from a raspberry pi running...
Read More →Deep learning has become essential for robots to operate and perform complex tasks in unfamiliar or dynamic environments. However, infrastructure and tooling for robot learning is largely insufficient–especially for integrated testing and evaluation of model performance in the real-world. To solve this problem, we’re working on an MLOps Suite for...
Read More →In our last update, we stood up Spring Boot microservices that our ESP32-S3 microcontrollers reach out to for model updates before downloading the new model. This sets us up nicely for flexible deploys of PyTorch models, and our next step is to implement the model runtime infrastructure to execute the...
Read More →As of our last update, we have already created pipelines that track a github repository, pick up PyTorch model changes, and export a model to an edge-device compatible format. That left us with a crucial question: what do we do next with the output artifacts of our pipelines?
Read More →Machine learning at the edge is steadily becoming more appealing for fast, secure, and reliable intelligence. At the onset of a robotics and IOT revolution, developers are wanting machine learning that is not dependent on some far-off server that brings with it the pain of rate/capacity limitations, lack of data...
Read More →