The Silicon Labs SiWx917 ultra-low-power Wi-Fi 6 product family has reached general availability (GA) for its AI/ML hardware accelerator. Accompanied with a guided machine learning (ML) development journey and other tools, the SiWx917 makes artificial intelligence (AI) on IoT (AIoT) accessible for everyone and speeds up edge AI transformation for Wi-Fi IoT device makers.

Edge AI Transformation

The world has become AI-driven. With smarter algorithms that go far beyond what traditional rule-based programming can ever achieve, AI improves accuracy, enhances functionalities, and enables new use-cases on almost any imaginable thing, function, and system. Anything that benefits from the advantages of AI is being enhanced with that capability as we speak. However, behind the scenes of this profound technological revolution, a significant transformation is taking place. While cloud computing has dominated the AI/ML landscape for years, the centralized model is now challenged with a distributed architecture that also operates at the network edges as well as on the IoT devices. Edge AI is a fundamental transformation driven by clear, quantifiable benefits.

The Benefits of Edge AI

While cloud computing has proven effective in making AI ubiquitous, it has faced significant challenges. The latency of sending data to the cloud and back does not scale for real-time applications. Processing and storing sensitive data on cloud servers raises privacy and security concerns. Vast amounts of data are causing constant network congestion, flooding cloud storage, and increasing energy consumption and CO2 emissions.

Edge AI distributes computing to billions of nodes. ML models run directly on IoT devices, close to where the data is generated, avoiding the need to send every bit of data to the cloud. The proliferation of IoT connectivity, rapid advancements in AI/ML hardware acceleration on wireless microcontrollers like the SiWx917, combined with easy ML development tools have made it possible to perform sophisticated ML computations directly on compact, resource-constrained devices.

With on-device AI/ML execution, billions of smart home appliances, IoT sensors, smart cameras, industrial equipment, and many other smart devices can now process data and make decisions autonomously, without relying on the cloud. The result is faster processing, real-time response, resilient applications, and reduced load on networks and data storage. Data stays where it was generated, improving privacy and security.

SiWG917 AI/ML Accelerator and Development Tools

The SiWG917 (part of the SiWx917 product family) is an ultra-low-power Wi-Fi 6 and Bluetooth Low Energy wireless MCU for energy-efficient and battery-powered IoT devices. The comprehensive Edge AI-capable system features Matter protocol support, a dual-processor architecture with a powerful application microcontroller, network wireless processor (NWP), and large SRAM, PSRAM, and flash memory capacity. Its rich set of peripherals and high GPIO count make SiWG917 a versatile wireless MCU solution for any smart IoT application.

The dedicated AI/ML hardware accelerator on SiWG917 is optimized for energy-efficient ML inference, offloading the Arm Cortex-M4 application MCU. The on-device optimized AI/ML inference capabilities include executing time series data processing ML models such as predictive maintenance, environmental monitoring, anomaly detection, voice and audio detection, and low-resolution vision use-cases.

The dedicated AI/ML hardware block on the SiWG917 is optimized for energy-efficient TinyML workloads. It can perform 320 MOPs on the dedicated accelerator, offloading this load from the Cortex-M4 microcontroller, and allowing it to process the application. This makes the SiWG917 a game-changer for energy-efficient, battery-powered Edge AI devices.

Edge AI is Now Accessible to IoT Developers

To make AIoT accessible to every developer, the SiWG917 AI/ML accelerator comes with an AI/ML Extension for Simplicity SDK that ensures a seamless use of the accelerator through a standard CMSIS-NN interface. The ML SDK includes TensorFlow Lite Micro (LiteRT) optimization for microcontrollers, running compact, efficient ML models on the IoT device.

Our online AI/ML developer journey guides you through the end-to-end development flow with sample apps, tooling, and a partner list, further lowering the bar for Edge AI innovation. Whether you’re a seasoned embedded engineer or a data scientist exploring the edge, the SiWG917 offers a seamless path from ML model training to deployment.

Internet access is essential to the productivity of most organizations—but not everything on the Internet is work-appropriate or safe. From time-wasting sites to phishing domains, employees can unintentionally expose your company to productivity loss, legal risks, or cyber threats just by visiting the wrong webpage. And before you think it couldn’t happen to you or your team—even the most cybersecurity aware can be duped by emerging threats like fake CAPTCHAs or phony unsubscribe links.

That’s why blocking unwanted websites on your company network is a critical step in your cybersecurity strategy. But how can you do it without over-restricting employees or overloading your IT team?

Let’s explore the most effective and scalable ways to block unwanted websites—without the headaches.

Why Should You Block Websites on Your Company’s Network?

Blocking specific websites isn’t about micromanaging employees—it’s about managing risk. Here are just a few reasons to implement web filtering:

• Security: Perhaps the most obvious reason, and undoubtedly one of the most important. Security-focused web filtering prevents access to suspicious domains, phishing sites, and malware-laden URLs.
• Compliance: Web filtering helps to enforce Internet usage policies and meet regulatory requirements like HIPAA, CIPA, or ISO27001.
• Productivity: Reduce distractions by filtering content like social media, video streaming, and gaming during work hours.
• Bandwidth Control: Limit access to high-bandwidth sites that can slow down business-critical applications.

Steps to Block Unwanted Content on Your Network

1. Identify What You Want to Block

Not all “unwanted” sites are created equal. Whether you only want to block threats, or you want to have stricter restrictions on employee Internet access, it’s important to define which categories or URLs pose a risk to your organization. Commonly blocked content categories include:

• Adult Content
• Gambling
• Streaming Media
• Social Networking
• Malware
• Phishing
• New Domains
  

The above is not a comprehensive list of content and threat categories to block, but a great place to start thinking about the types of content you’d like to limit access to. Find a more complete list of content category options on the DNSFilter Help Center.

2. Choose the Right Website Blocking Method

There are multiple ways to block websites on a company network. Here are the most common approaches:

DNS Filtering (Recommended)

DNS filtering blocks access to unwanted websites at the DNS level—before the connection is even established.

Benefits:

• Lightweight and fast
• Works across all devices on your network
• On-device software is not required
• Cloud-based and easy to scale

Firewall Rules (Limited Use Case)

Traditional firewalls can block domains or IP addresses, but they often lack the context or categorization needed for nuanced filtering.

Browser Extensions or Proxy Servers

These can offer site blocking on specific devices but require extensive setup and maintenance. They’re also easier for savvy users to bypass.

3. Create Flexible Policies

Web content filtering isn’t one-size-fits-all. Different departments might have different needs. For example, your marketing team might need access to social media while the accounting department doesn’t.

With DNSFilter, you can create custom filtering policies to block employee access to content by:

• Group or department
• Device or user
• Time of day (e.g., block streaming during work hours only)
• Network location
  

Beyond just filtering by content or threat category, Allow and Block lists allow you to create granular policies down to the domain, subdomain, or TLD.

This customization allows you to strike the perfect balance between security and productivity.

4. Monitor and Adjust

Blocking sites isn’t a “set it and forget it” task. Threats evolve, and so do your employees’ needs.

Use your content filter’s analytics and reporting tools to monitor:

• Attempts to access blocked content
• Emerging threats or high-risk domains
• Bandwidth usage by category

With this insight, you can fine-tune your policies and stay one step ahead of risks.

See how easy it is to create content filtering policies in a few minutes using DNSFilter. Try the interactive product tour →

5. Communicate the Policy

It’s important that employees understand why certain sites are blocked. Transparency builds trust and helps prevent accidental policy violations. Additionally, it’s important to give your users a way to submit requests to access content that has been unnecessarily or incorrectly blocked.

Ensure that feedback is a two-way street between your Security team and the rest of the company’s users. Provide a clear Acceptable Use Policy and include messaging in customizable block pages to let users know when and why a site is restricted, and also how to submit their requests for a site to be unblocked or recategorized.

DNS Filtering is Key for Blocking Unwanted Websites

Blocking unwanted websites doesn’t have to mean locking down your network or frustrating your users. With intelligent, AI-powered DNS filtering, you can protect your business from threats, enforce policies, and ensure a more productive digital environment.

Ready to take control of your network? Try DNSFilter free for 14 days and see how easy it is to block the bad without breaking your business.

Set-top boxes and related products can be enhanced with leading AI-native Edge processors, contextual awareness, and reliable wireless connectivity to create enriched viewing experiences.

With the growing adoption of IoT devices and increasing demand for video surveillance, enterprises expect more from remote monitoring. The standard is no longer just “connectivity” or “low power consumption,” but a combination of long-range coverage, stability, low latency, and high reliability. Built on the IEEE 802.11ah standard, Wi-Fi HaLow is designed for long-range, low-power applications and is quickly emerging as a strong choice for remote video monitoring.

Core Challenges in Remote Video Monitoring

Traditionally, remote monitoring required a trade-off between high-bandwidth Wi-Fi and highly power-efficient LPWAN technologies. Conventional Wi-Fi is widespread and convenient, but its high power consumption and limited coverage make it less effective for large-scale monitoring scenarios. As coverage areas expand, equipment and cabling costs rise significantly, along with the burden of ongoing maintenance.

On the other hand, LPWAN technologies offer excellent long-range and low-power capabilities, but their narrow bandwidth restricts usage to low-data-rate sensor applications, falling short of supporting high-definition video transmission.

When both high bandwidth and low power are required, enterprises have typically relied on complex hybrid solutions. Wi-Fi HaLow changes this equation by delivering long-range communication, low-power operation, and higher throughput in a single technology, enabling the seamless integration of sensor data with real-time video for efficient and reliable wide-area monitoring.

Key Features of HaLow in Remote Video Monitoring

Extended range and penetration: Operating in the sub-1GHz spectrum, HaLow signals penetrate walls more effectively and maintain stable coverage in open outdoor spaces or environments with heavy metal structures.

Low-power operation: Energy-efficient modes allow cameras and IoT devices to run for extended periods, reducing power consumption and maintenance costs.

Scalable network architecture: HaLow supports MESH networking, enabling large-scale device deployments with dynamic routing to ensure stable video transmission.

Cost and traffic efficiency: Wi-Fi HaLow requires no additional subscription fees. Its MESH architecture enables the creation of large-scale private networks, minimizing cloud dependency, reducing communication costs, and enhancing data security.

Transformations with Wi-Fi HaLow

The adoption of Wi-Fi HaLow brings tangible improvements to remote monitoring scenarios. First, HaLow unifies network requirements across different communication devices, eliminating the need for complex gateways or multi-protocol endpoints, which simplifies deployment and increases efficiency.

Second, HaLow strikes a balance between traditional Wi-Fi and LPWAN strengths, reducing the cost and complexity of deploying and maintaining large-scale networks. With fewer devices, users can cover larger areas while still benefiting from higher bandwidth to ensure stable video and data transmission.

Additionally, HaLow supports up to 8,191 client connections per access point, moving beyond traditional point-to-point limitations. This scalability allows thousands of IoT devices to connect simultaneously, making it ideal for high-density environments such as smart factories, industrial parks, or city-wide surveillance systems.

Finally, Wi-Fi HaLow eliminates ongoing data subscription fees. Combined with MESH networking, it enables large-scale private networks where information does not need to be routed entirely through the cloud, improving security while further cutting communication costs.

Conclusion

Wi-Fi HaLow not only extends IoT device communication range but also upgrades remote monitoring from simple data collection to comprehensive, visualized, intelligent surveillance systems. From remote farms to industrial parks, from city safety to large-scale warehouses, HaLow enables reliable transmission of both high-definition video and sensor data, setting a new standard for remote monitoring.

AsiaRF provides a complete portfolio of HaLow solutions, including some of the market’s rare HaLow MESH networking devices and the world’s first plug-and-play USB HaLow dongle, which significantly lowers deployment costs and technical barriers. This allows enterprises and developers to easily adopt HaLow and benefit from long-range, low-power, and reliable remote monitoring at scale.

Wi-Fi Offloading, How? – Chapter 2.1

White Paper: Wi-Fi Offloading, How?

This is an excerpt from our white paper, Wi-Fi Offloading, How?,  a technical deep dive into deploying Wi-Fi offloading solutions. If you like what you read, download the full white paper. As a bonus, you’ll also gain access to Wi-Fi Offloading, Why?, outlining the business benefits for mobile operators.

DOWNLOAD    INSIGHTS    CONTACT US

Optimizing AI performance on Synaptics’ Astra™ platform with extreme low-bit quantization.

As AI continues to move from the cloud into everyday devices, the ability to run models efficiently on the Edge is becoming increasingly important. Whether it’s voice interfaces or real-time data processing, Edge AI promises a wide range of capabilities. Delivering those capabilities within the constraints of embedded systems, however, remains a challenge.

ENERZAi has partnered with Synaptics to address these challenges. Known for their advanced Edge processing platforms, Synaptics provides a foundation for deploying optimized AI models. Together, we’re focused on making high-performance AI more practical for real-world Edge applications.

Making AI Inference Lighter and More Efficient

ENERZAi is focused on improving inference performance through model compression and optimization. Our software engine, Optimium, is designed to run trained models on devices with limited compute, memory, and power. A key part of this approach is extreme low-bit quantization. While many AI systems use 8-bit or 4-bit quantization to reduce model size, our method reduces to just 1.58 bits. This allows for significantly smaller models and faster inference.

Deploying Whisper on the Synaptics Astra SL1680 Platform

In our work with Synaptics, we applied 1.58-bit quantization to OpenAI’s Whisper small model and deployed it on the Astra SL1680 processor. With its quad-core 2.1GHz Arm® Cortex®-A73, Astra provides the right balance of compute and efficiency for Edge AI applications.

The results highlighted how optimized inference and advanced quantization can work together:

• The quantized model achieved a Word Error Rate (WER) of 6.38 percent, compared to 5.99 percent for the FP16 baseline
• 4x reduction in peak memory usage compared to FP16
• 2x Inference latency reduction for a 9-second audio input as compared to the full-precision version

These gains are significant for real-world Edge applications, enhancing system stability and user experience, especially in environments where multiple AI workloads need to run in parallel.

Partnering to Advance AI at the Edge

Synaptics and ENERZAi’s partnership advances Edge AI, combining compression technology with the robust capabilities of the Optimium engine. The versatile CPU, GPU, and NPU subsystems within the Astra SL1680 make Edge AI more responsive, efficient, and deployable across a range of applications.

For more details, read the full solutions brief on Running Extreme Low-Bit Models on IoT Edge Devices here:  Running_Extreme_Low-Bit_Models_on_IoT_Edge_Devices_4.pdf 

World’s first fully integrated RAIN RFID reader in an enterprise mobile processor offers compact form factor, low BOM cost, smaller footprint, end-to-end commercial readiness and superior performance.