Since 2019 we have been involved with the design, development, and deployment of AI-based solutions for our customers and with our partners. Some examples are illustrated below, followed by a short description.
Building a machine learning-centric PKM with Obsidian
This ongoing project will leverage large AI models with knowledge graphs based on Obsidian notes.
Figure: Auto-generated semantic graph
Knowledge graphs and queries are based on semantic information that augments the Obsidian graph.
OnestepAI: AutoML for IoT
Figure: OSAI Portal
We consulted on the design and promotion of a cloud-hosted AutoML solution for people with little AI knowledge. Complete cloud solutions from uploading your image dataset, to training a model to running inferences on cloud-hosted RPI and Jetson Nano boards.
Alftel: Coral EdgeTPU x 12 board
Figure: PCIe gen3 with 12 Coral EdgeTPU on-board
Sourced from Alftel
Custom Coral EdgeTPU (12x) board used to process at very high speed a combination of AI models and remote video feeds.
Mininodes: ARM64 with Gyrfalcon
Figure: RPI4 server with Gyrfalcon AI accelerator
Precursor to OSAI for the cloud hosting of RPI4 boards equipped with a Gyrfalcon USB based accelerator.
Figure: NanoPI server with Gyrfalcon AI accelerator
The smallest AI server!
Figure: Inference server
Example of a custom-trained classifier with a remote web app front-end hosted on an anvil served front-end.
Solidrun: Honeycomb with Gyrfalcon AI accelerator
Figure: CES2020 demo
An ARM64 server with 16 x A72 cores with Gyrfalcon accelerators. Was used in a real-time TV feed streaming application.
Robosensing: Various AI projects
Figure: Solar-powered AI camera and servers
Combining the small form factor and power requirements of an RPI board with a high-speed AI accelerator. Solar energy was enough to power the entire setup.
Figure: BentoAI leaning camera
Few-shot learning application for rapid registering of several new objects’ class.
Figure: CES2020 demo
End-to-end demo
Figure: CES2020 demo
End-to-end demo
Figure: CES2020 demo
RPI4-based Kubernetes cluster equipped with Gyrfalcon AI accelerators.
