BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
He too will find the right place to sniff around.
We usually use a lamp post for support.
#hiring | BrainChip
We're #hiring a new AI Hardware Design Engineer (MPU) in Laguna Hills, California. Apply today or share this post with your network.
TheDrooben
Pretty Pretty Pretty Pretty Good
AI's Hidden Constraint: Power Efficiency and Edge Computing | Doug Utberg posted on the topic | LinkedIn
AI is hitting a constraint no one wants to talk about: power. Most conversations fixate on model size, GPU clusters, and trillion-dollar data center builds. But thereâs a physical bottleneck emerging. In my latest conversation with Sean Hehir, CEO of BrainChip, we unpack why AIâs future isnât...
Happy as Kevin
Last edited:
Thank you for your prompt reply. I will look for it and buy two. I will continue to follow BRN indefinitely. Only when the SP touches the upper chart line at 7 will I give my wife a shirt.
Ehhh why should I know if you will give your wife a shirt?
Donât get itâŠ
Letâs see who will be triggered
Two new oemâs
trustedsemi.com
https://brainchip.com/partners/
Trusted High Reliability Microelectronics | Electronics development | Solutions for High Reliability Microelectronics
Trusted Semiconductor Solutions is a high reliability semiconductor design and product development company with expertise in radiation hardened microelectronics.
trustedsemi.com
https://brainchip.com/partners/
Last edited:
Frangipani
Top 20
As suspected, the recent articlethat claimed BrainChip âhad secured an order from Nex Novus for medical sensing applicationsâ was wrong:
There will be a live demo of the made-in-Croatia Neuromorphyx Vision NeuroNode at the BrainChip booth at Embedded World 2026 in NĂŒrnberg (10â12 March).
âExperience the Vision NeuroNodeâą tiny, compact and rugged form in-person! Featuring the AKD1500 neuromorphic co-processor from Brainchip Inc. pushing the boundaries of high-performance, low-latency, Spiking Neural Networks (SNN) for energy efficient edge AI.
POWER & COMPUTE EFFICIENCY < 300 mW while delivering 800 GOPS
ON-CHIP LOCAL MEMORY 1 MB
CLOCK FREQUENCY RANGE 5 â 400 MHz
SILICON PROCESS 22 nm FD-SOI CMOS digital logic
PACKAGE 7Ă7 mm MFCTFBGA169, 0.5 mm pitch
Find us at:
Hall 5 / Booth 5-213 (10 - 12 March 2026. Nuremberg, Germany)
NeuroNodeâą
Deployment time - seconds.
Operational lifetime - years.
Scalability factor - unparalleled.
Rugged build.
Always-on sensing.
Energy-efficient processing.
Battery-powered operation.
Low-latency reactions.
The Vision NeuroNodeâą is armed with an FPGA, a neuromorphic AI accelerator and an event-based neuromorphic vision sensor.
Our IP and architecture, allows up to multiple-years* of field operation on a single battery pack integrated in our compact 5x5x5cm IP67 case.
*Depending on scene activity and inference.
We offer solar panels and add-on battery packs as attachments that further prolong the operational life-time and extend capabilities.
Technology developed by Prophesee and Sony.
- Fast tracking, equivalent to >10.000 FPS
- Extreme lightning conditions >120dB
- Native power efficiency <10mW
Skillfully engineered into compact modules by CenturyArks.
We offer a wide range of sensors and lenses to meet our client requirements depending on deployment location, terrain and field conditions.
HD 1280 x 720 Sony IMX636 / IMX646 (via CenturyArks)
VGA 640 x 512 Sony IMX 637 / 647 (via CenturyArks)
320 x 320 ultra low-power GENX320 (via Prophesee)
Explore sensor features in the Sensor options section ââ
[âŠ]
NeuroHiveâą
Vision, Audio, Radar, IMU are different NeuroNodesâą forming the hive mind network. Distributed intelligence, at the edge.
NeuroHiveâą provides your team with the ability to gather mission critical intel from the field, live. Scalable and reliable.
Over-the-air SNN model switching.
Tracking target velocity and direction.
Fleet management.
Orchestrated systems.
External API triggers.
Integration to existing systems and comms.â
Neuromorphyxâą - Edge AI Networks
Networks of neuromorphic edge AI nodes. Distributed by hardware. Unified by software. Sense. Think. Decide. On the edge. NeuroNodesâą managed via NeuroHiveâą.neuromorphyx.com
View attachment 95475 View attachment 95476 View attachment 95477 View attachment 95478 View attachment 95479 View attachment 95480 View attachment 95481 View attachment 95483
View attachment 95482
The Neuromorphyx LinkedIn account is now active:
Saw this earlier todayâŠ
#neuromorphic #embeddedworld #edgeai | Neuromorphyx
Neuromorphyx is live. Edge AI nodes. Distributed Awareness. Stay tuned. #Neuromorphic #EmbeddedWorld #EdgeAI
Meanwhile there has been another post, this time mentioning BrainChip:
NeuroNode | Neuromorphyx
The "Vision" NeuroNodeâą - Made in Croatia, EU đȘđș đđ· Join us at BrainChip booth 5-213 / hall 5 - embedded world Exhibition&Conference 10 â 12 Mar 2026, NĂŒrnberg. Featuring event-based sensors: PROPHESEE and CenturyArks Co., Ltd. Neuromorphic co-processor / accelerator: BrainChip FPGA...
LinkedIn Login, Sign in | LinkedIn
Login to LinkedIn to keep in touch with people you know, share ideas, and build your career.
I also noticed the Nex Novus website looks different than a month ago:
As Terroni had already shared back in January, the founder and CEO of Nex Novus - and also Co-Founder and CEO of Neuromorphyx - is Ivan ProjiÄ, who resides in Copenhagen but has registered the business in his native Croatia, in Pula, where his Neuromorphyx Co-Founder and CTO Filip Gembec lives:
Last edited:
Frangipani
Top 20
Jean-Luc Chatelain, Founder and Managing Partner at Verax Capital Advisors as well as former MD and Global CTO at Accenture Applied Intelligence has been a a long-time friend of BrainChipâs. Many forum readers will be familiar with his name thanks to him getting quoted on the BRN website 
or because theyâve listened to a podcast with him.
Earlier today, he reposted a BRN job ad for better reach and wrote
âThis is a great opportunity for one who believes, as I do, that neuromorphic computing will have significant contribution to the future of #AI at scaleâ.
www.linkedin.com
brainchip.com
or because theyâve listened to a podcast with him.Earlier today, he reposted a BRN job ad for better reach and wrote
âThis is a great opportunity for one who believes, as I do, that neuromorphic computing will have significant contribution to the future of #AI at scaleâ.
#ai | Jean-Luc Chatelain
Reposting for amplification. This is a great opportunity for one who believes, as I do, that neuromorphic computing will have significant contribution to the future of #AI at scale
Home
Unlock the power of AI with BrainChip. Enhance data processing, Edge apps and neural networks at the speed of tomorrow. Explore now!
brainchip.com
smoothsailing18
Regular
FF
François Piednoel de Normandie 3rd+
Athos Silicon Cofounder, ex-Performance Gurus of Glorious Intel. Ex-Mercedes Benz ADAS hardware Architect. IEEE Member
2w Edited
Follow
One of the patents I am most proud of in autonomous driving is now pending.To understand it, you first need to understand the statistical view of the world described in the DOGMA paper, authored by my former coworkers:https://lnkd.in/g8tHnar9Once you grasp this, you understand the fundamentals of how to build a system that provides the autonomous-driving stack with a statistical model of the environment around the car, allowing you to quantify how reliable the statistics are for every single volumetric cell (âcubeâ) in Î-space around the vehicle.This approach is extremely powerful for sensor fusion: combining multiple LiDARs, radars, and cameras while explicitly reasoning about the reliability of each measurement. It can significantly reduce well-known side effects of LiDARs and radars such as highly reflective surfaces because those measurements are statistically weak and therefore down-weighted.However, this typically requires a large and expensive LiDAR setup, like the one used by Waymo.The new pending patent takes a different approach. It uses a set of fixed, single-point LiDARs, roughly $2 devices. When arranged properly, these sensors re-increase the statistical confidence along entire lines of volumetric cells by introducing additional independent measurements. This dramatically reduces the number of expensive LiDAR units needed by replacing them with far more affordable ones, while preserving statistical robustness.And yes, the patent also references an EQXX-like shape, the current record holder for EV range:https://lnkd.in/gg4QnXJn(I am genuinely in love with this car, and I hope Mercedes-Benz will one day produce it commercially.)Importantly, this sensor fusion is not intended to directly drive the car. Its purpose is to verify the trajectories proposed by the machine-learning driving stack and to reject any trajectory that does not meet a confidence threshold of 1 chance in 100 million of being wrong (100Ă ISO 26262).To perform this work end-to-end, you need very special hardware. A GPU is not suitable here: burning 300 W while carrying severe safety limitations makes it fundamentally incompatible with this level of reliability.The full patent portfolio will be published in 2026.I hope youâll enjoy the ride.
34
6 Comments
2 Reposts
Heiner Stockmanns
Semiconductor Executive â retired and having fun
2w
Keep going and I cheer you on. Thoroughly enjoying folks tackling extremely challenging problems while striving for ultimate safety and affordability. Never stop differentiating yourself from the pretenders.
Like
Reply
1 Reaction
Andrew Bullen
Retired - BlackBerry Shareholder QNX Supporter
2w
François Piednoel de Normandie - very cool !!!
Like
Reply
Kieran Ryan
Medical Device Expert. Trainer, Inventor, Market Developer and Clinical Cover!
2w
Can you utilise other hardware yet which utilises less power? Maybe something neuromorphic?
Like
Reply
François Piednoel de Normandie
Athos Silicon Cofounder, ex-Performance Gurus of Glorious Intel. Ex-Mercedes Benz ADAS hardware Architect. IEEE Member
2w
Kieran Ryan Yes , there is an hardware scheduler
Like
Kieran Ryan
Medical Device Expert. Trainer, Inventor, Market Developer and Clinical Cover!
2w
François Piednoel de Normandie so the stack works independent of the processor controlling it. Thatâs very helpful
Like
François Piednoel de Normandie
Athos Silicon Cofounder, ex-Performance Gurus of Glorious Intel. Ex-Mercedes Benz ADAS hardware Architect. IEEE Member
2w
Kieran Ryan , no, we avoid any non-deterministic solution, as it is what is used to certify the safety side of our hardware + software.
Like
1 Reaction
(Reposted by a neuromorphic computing engineer at Mercedes Benz)
François Piednoel de Normandie 3rd+
Athos Silicon Cofounder, ex-Performance Gurus of Glorious Intel. Ex-Mercedes Benz ADAS hardware Architect. IEEE Member
2w Edited
Follow
One of the patents I am most proud of in autonomous driving is now pending.To understand it, you first need to understand the statistical view of the world described in the DOGMA paper, authored by my former coworkers:https://lnkd.in/g8tHnar9Once you grasp this, you understand the fundamentals of how to build a system that provides the autonomous-driving stack with a statistical model of the environment around the car, allowing you to quantify how reliable the statistics are for every single volumetric cell (âcubeâ) in Î-space around the vehicle.This approach is extremely powerful for sensor fusion: combining multiple LiDARs, radars, and cameras while explicitly reasoning about the reliability of each measurement. It can significantly reduce well-known side effects of LiDARs and radars such as highly reflective surfaces because those measurements are statistically weak and therefore down-weighted.However, this typically requires a large and expensive LiDAR setup, like the one used by Waymo.The new pending patent takes a different approach. It uses a set of fixed, single-point LiDARs, roughly $2 devices. When arranged properly, these sensors re-increase the statistical confidence along entire lines of volumetric cells by introducing additional independent measurements. This dramatically reduces the number of expensive LiDAR units needed by replacing them with far more affordable ones, while preserving statistical robustness.And yes, the patent also references an EQXX-like shape, the current record holder for EV range:https://lnkd.in/gg4QnXJn(I am genuinely in love with this car, and I hope Mercedes-Benz will one day produce it commercially.)Importantly, this sensor fusion is not intended to directly drive the car. Its purpose is to verify the trajectories proposed by the machine-learning driving stack and to reject any trajectory that does not meet a confidence threshold of 1 chance in 100 million of being wrong (100Ă ISO 26262).To perform this work end-to-end, you need very special hardware. A GPU is not suitable here: burning 300 W while carrying severe safety limitations makes it fundamentally incompatible with this level of reliability.The full patent portfolio will be published in 2026.I hope youâll enjoy the ride.
34
6 Comments
2 Reposts
Heiner Stockmanns
Semiconductor Executive â retired and having fun
2w
Keep going and I cheer you on. Thoroughly enjoying folks tackling extremely challenging problems while striving for ultimate safety and affordability. Never stop differentiating yourself from the pretenders.
Like
Reply
1 Reaction
Andrew Bullen
Retired - BlackBerry Shareholder QNX Supporter
2w
François Piednoel de Normandie - very cool !!!
Like
Reply
Kieran Ryan
Medical Device Expert. Trainer, Inventor, Market Developer and Clinical Cover!
2w
Can you utilise other hardware yet which utilises less power? Maybe something neuromorphic?
Like
Reply
François Piednoel de Normandie
Athos Silicon Cofounder, ex-Performance Gurus of Glorious Intel. Ex-Mercedes Benz ADAS hardware Architect. IEEE Member
2w
Kieran Ryan Yes , there is an hardware scheduler
Like
Kieran Ryan
Medical Device Expert. Trainer, Inventor, Market Developer and Clinical Cover!
2w
François Piednoel de Normandie so the stack works independent of the processor controlling it. Thatâs very helpful
Like
François Piednoel de Normandie
Athos Silicon Cofounder, ex-Performance Gurus of Glorious Intel. Ex-Mercedes Benz ADAS hardware Architect. IEEE Member
2w
Kieran Ryan , no, we avoid any non-deterministic solution, as it is what is used to certify the safety side of our hardware + software.
Like
1 Reaction
(Reposted by a neuromorphic computing engineer at Mercedes Benz)
buena suerte :-)
BOB Bank of Brainchip
Bravo
Meow Meow đŸ
Hi Smoothy,
It looks like the comments may not be sequenced correctly in your post.
From what I can see (screenshot below), when Kieran Ryan asks: "Can you utilise other hardware yet which utilises less power? Maybe something neuromorphic?"
Piednoel responds "No, we avoid any non-deterministic solution, as it is what is used to certify the safety side of our hardware + software. "
Out of curiosity, I asked ChatGPT what it made of that exchange. In summary, it suggested:
- Neuromorphic architectures are generally considered non-deterministic in behaviour (at least from a safety certification perspective).
- At this point in time neuromorphic architectures don't meet automotive safety standards (ISO 26262, ASIL) prioritise deterministic, repeatable execution paths.
- If the system is part of a safety-certified validation layer, engineers would typically avoid architectures that are difficult to formally verify.
- His mention of a âhardware schedulerâ suggests a deterministic co-processor or ASIC approach rather than a neuromorphic one.
I think ChatGPT might be being a bit conservative about the 5-7 year estimate for automotive adoption since Mercedes have commented previously, unless I'm mistaken, that they're looking at 2030 time-frame.
Happy to hear alternative interpretations, but thatâs how it reads to me.
smoothsailing18
Regular
Hi Bravo...... it's all way over my pay grade, cut and paste only.
However I noticed the below was the interesting part for myself.
Like
1 Reaction
(Reposted by a neuromorphic computing engineer at Mercedes Benz)
However I noticed the below was the interesting part for myself.
Like
1 Reaction
(Reposted by a neuromorphic computing engineer at Mercedes Benz)
This sure sounds like Akida. Any links to us???
Type 1 Compute | Brain-inspired processors for edge AI
Extract :-â
The Event-Driven Advantage
Traditional cameras capture full frames 30 times per secondâwasting power processing unchanged pixels. Event-driven sensors fire only when pixels detect change, enabling:
SPACE SURVEILLANCE: Track satellites moving 17,000 mph against bright Earth or dark space simultaneouslyâimpossible for frame cameras that saturate or lose contrast.
This same principle scales to:
Type 1 Compute | Brain-inspired processors for edge AI
Extract :-â
The Event-Driven Advantage
Traditional cameras capture full frames 30 times per secondâwasting power processing unchanged pixels. Event-driven sensors fire only when pixels detect change, enabling:
SPACE SURVEILLANCE: Track satellites moving 17,000 mph against bright Earth or dark space simultaneouslyâimpossible for frame cameras that saturate or lose contrast.
This same principle scales to:
- âąAutonomous vehicles navigating at 45+ mph
- âąIndustrial systems detecting vibration frequencies
- âąDrones adapting to their environment in real-time
- âąRobotics requiring instant obstacle detection
Similar threads
