Eight cameras, one $249 board
Object detection across eight independent streams, four live and four recorded, tiled onto a single monitor.
8 streams · 1× Jetson Orin Nano · $249
running live
Why us
We get more out of less silicon
Everyone on this floor says edge AI. Running vision on a board with a big accelerator proves very little. The hardware absorbs the inefficiency. Running the same workload on a part with no NPU at all takes real quantization and cycle-budget engineering. That's the range we work.
EdgeCloud
- 01
Microcontroller
64 KB – 2 MB, no NPU
- 02
Embedded Linux
512 MB – 2 GB
- 03
Edge accelerator
up to 67 TOPS
- 04
On-prem server
GPU
- 05
Cloud
elastic
At the booth
What's running
Four to look at first. Everything is on real hardware in front of you, not a video of something that once worked.
Defect detection with a reject arm
PCB inspection on a moving belt, with a 6-DOF arm pulling failed boards off the line. Inspect, classify and act, all on the device.
measurement added after the dry run
running live
People counting without a camera
60 GHz FMCW radar tracking occupancy and direction through a doorway. No image data leaves the sensor, because there is none.
measurement added after the dry run
running live
Describing a defect it was never trained on
A vision-language model narrating faults on the same boards as the trained detector, run side by side so you can see where each one wins.
measurement added after the dry run
on request
There are more demos than there is bench space, so a subset runs live at any moment and the rest sit warm. Ask for anything you don't see on the bench and we'll have it up in a minute or two.
Book a slot
Tell us which demo, and roughly when
We'll keep it hot and make sure an engineer is free rather than mid-conversation.
info@brillersys.com
Goes to the team inbox, not a CRM sequence.
The booth is four days of it. The other fifty-one weeks are six practices running from microcontrollers to cloud platforms.
See our services →