Mike Fine

A Utah school had no usable cellular service inside. Calls dropped. Texts failed. Emergency calls didn’t go through. InfiniG partnered with AT&T and T-Mobile to bring their signal indoors - delivering reliable voice and data coverage in every hallway and classroom. Now staff, students, and parents stay connected when it matters most.

79

2 Comments

And Utah releasing results tonight as well. Utah is proposing to utilize $231M (73%) of its eligible BEAD funding to serve 32,261 eligible locations and CAIs. The technology mix is fiber (55%), LEO (25%), LFW (17%) and ULFW (2%). #BEAD #RuralBroadband

27

Imagine being responsible for 600 students, and your phone won’t connect to 911. That was the situation at Parkside Elementary in Utah. Modern building materials and a valley location left teachers without signal for emergency calls. The traditional solution? A repeater—but there was no external signal to amplify. Our solution? Mobile Coverage-as-a-Service: think “cellular as a service.” We utilize small-cell radios and cloud software, making it subscription-based and deployable within months, supporting all carriers (neutral host). When Parkside approached InfiniG: In just 4 months, reliable indoor coverage was established. - 911 calls and panic buttons now function instantly. - Parents, staff, and students stay connected. - The system is future-proofed for tools like AI in classrooms. This wasn't merely an upgrade; it transformed safety. Because if 911 doesn't work indoors, nothing else matters.

32

I find Greg McKeown’s Essentialism: The Disciplined Pursuit of Less to be a transformative read for anyone seeking clarity and impact in their work and life. McKeown challenges the pervasive “more is better” mindset, urging readers to focus on the “vital few” rather than the “trivial many”. The core message is clear: success and fulfillment come not from doing more, but from doing what truly matters. The book offers a systematic approach—identify your highest point of contribution, eliminate non-essentials, and make execution effortless. Essentialism isn’t about getting more things done; it’s about getting the right things done. McKeown’s practical strategies and real-world examples make the philosophy actionable for professionals and leaders alike. A powerful takeaway: “If you don’t prioritize your life, someone else will.” By embracing trade-offs and learning to say no, we can reclaim our time, energy, and purpose. Essentialism is a must-read for anyone ready to break free from overwhelm and operate at their highest potential.

51

4 Comments

Paralyzed by too much attention to detail... a story. (This story is a little long, but I promise it has value).   Back around 2001, I was Director of National System Development for Western Wireless (Cellular One). At HQ in Bellevue, WA, it came to my attention that we had audio boards for Lucent cellular switches unevenly distributed across our markets. For example, one Mobile Telephone Switching Office (MTSO) in some state might have ten of those boards in their antistatic wrappers on a shelf in the back room. Meanwhile, in another state, the GM of the MTSO there might have to place an overnight rush order for an audio board because he had no spares at all.   I issued one of those emails from on high: “All GMs at Lucent locations, count your audio boards and send me an email. I want to know how many you have in service, and how many spares you have.”   Time passed. All but one General Manager had reported their numbers. I called that GM and said, “How soon can you get me the count on your audio boards?” His reply made the reason for his reluctance clear.   He said, “To get the serial number on the audio boards in service, I have to take the system offline. You can’t read the number from the front of the rack. We don’t have any downtime maintenance scheduled right now, so I haven’t got the serial numbers yet.”   I clarified what I was asking for. “I don’t want to know the serial numbers. I didn’t ask for the serial numbers. I want to know how many you have in service, and whether you have an adequate number of spares on site.”   He answered immediately. He knew the answer. “Oh. Well, if that’s all you need, I have twelve in service and one spare. Please don’t take my spare.”   I replied, “Take your spare? Heck no, I’m going to send you another one!”   He was bewildered that I didn’t want the serial numbers. As we continued our conversation, I explained it this way: “Our company is a bucket, and the audio boards are grains of sand in the bucket. I want to know how much sand is in the bucket. I don’t want to know the serial number of each grain.”   THE LESSON Be intentional about the information you collect. A lot of time is spent on information collection and storage that will never provide value. Over the years, the waste caused by this activity has gotten worse, not better. Storage is cheap (by comparison to what it used to cost), and search and sort algorithms are efficient. As a result, the default position is to scoop up information because we never know when we might be able to use it or monetize it. The problem is that we’re hoarding data. You heard me. Without focus, without clarity, without direction, without a plan, that pile of data is indicative of a hoarding problem. The storage may be cheap, but the human resource cost is often higher than estimated. Get a vision. Get focus, clarity, and direction. And let go of everything else, because everything else is a distraction.   #CallMeIfYouNeedMe #FIFONetworks   #InformationTechnology

18

1 Comment

Every year at Eigenvector University in Seattle we have the Wednesday evening "PLS_Toolbox/Solo PowerUser Tips & Tricks" session. It's a chance for all us EigenGuys to show features in our software that we think are often overlooked or under-appreciated. This year I decided to talk about custom cross-validation (CV). Typically, as a default, cross-validation sets are formed with the "venetian blinds" method. For example, if using a 5-way split, test sets are formed with every 5th sample, starting from the first, second, third sample and so on. However, that way of cross-validating might not reflect the way in which a model is to be used and lead to over-optimistic estimates of how it will perform. As an example, consider the Chimiométrie Challenge data from 2018. The calibration data consists of the NIR spectra of 10 different foodstuffs, 3908 samples total. The goal is to predict the protein content. You can put all of this data into a single model, in this case a small neural network, and if you cross-validate it with venetian blinds you might think that you're getting close to having a universal protein predictor (top panel). The error of calibration (RMSEC) is ~1.5% and error of cross-validation (RMSECV) ~1.7%. But if you cross-validate by foodstuff class, where each foodstuff is left completely out and then the model attempts to predict it, you find the model doesn't work on foodstuffs that it has not seen before (bottom panel). Error of cross-validation is now ~7.8%! So the trick is to set up your cross-validation in a way that best mimics how the model is to be used. Like any exam, test sets should be tough but fair! And of course PLS_Toolbox and Solo make it easy to use class sets for cross-validation, or you can create your own CV, including samples that must always be in or out. #crossvalidation #regression #calibration #ann

109

6 Comments

💡 Radar People: remember when “getting a test target on the screen” meant a year-long spec doc, an avalanche of NRE, and a six-figure box that arrived just in time for your program review? 😏 Been there. Burned the budget. Fast-forward to MATLAB R2025a where a new shipping example, "Radar Target Emulation on NI USRP Radio", lets you spin up a multi-target emulator in an afternoon: * Open the Simulink model * Push HDL to the USRP’s FPGA * Run the auto-generated host script * Watch up to 4 virtual targets dance across your range-Doppler map — no black-box “magic hardware” required What used to cost 6 and 7 digit $$$ (and lots of gray hair) now boils down to: * An off-the-shelf NI USRP radio * MATLAB, Simulink & a handful of toolboxes * A pot of coffee ☕ * Some assembly required 😉 The workflow ships since R2025a and walks you from model to bitstream to live demo. Bottom line: if your radar lab still treats target emulation as an exotic procurement, it’s time for a re-think. Grab the example, grab a USRP, and build it yourself — before the next procurement board meeting. 🔗 Details here: https://lnkd.in/eNZBMMNM #TargetGenerator #Radar #SignalProcessing #MATLAB #Simulink #USRP #Prototyping

100

6 Comments

The Chippewa Cree Tribe of the Rocky Boy’s Reservation of Montana recently turned up a new broadband network leveraging Ciena’s broadband solution that provides fast, reliable, and affordable internet access to its tribal households, businesses, and public service entities, such as the health center, police department, and schools. Learn more at https://bit.ly/4l5cOSF

47

1 Comment

Half of emergency CTs shouldn’t even be done.  In a recent paper, researchers found that over 50% of emergency CT requests were inadequate or unjustified Of nearly 1,300 emergency CT requisitions: - 22.2% - adequate - 5.5% - barely adequate - 65.9% - considerably limited - 4.1% - deficient - 2.2% - lacking adequate information The reasons are quite obvious - and familiar to everyone: → Ordering physicians under pressure → Vague clinical indications → Fear of missing something → And the convenience of “just scan it” But every unnecessary CT comes at a cost: - Radiation exposure     - False positives that trigger cascades of further testing    Most importantly, they consume radiologists’ time - already a scarce resource, preventing them from focusing on the cases where imaging was actually needed. The authors propose a new classification, RI-RADS, that can provide immediate feedback on the justification of each radiology request. Do you think this approach could help? It might - but if it remains purely a recommendation, fear of missing a life-threatening diagnosis will probably keep the “just in case” scans coming. Team: Xianwei Liu Zhengguang Xiao Qinghua Min Tianhao Wu, Yue Xing Yangfan Hu Defang Ding Shun Dai Junjie Lu Jiarui Yang Yue Li Yang Song Minda Lu Jingshen Chu Huan Zhang Weiwu Yao Jingyu Zhong  Link: https://lnkd.in/gH8RVcpv

120

49 Comments

A year ago we were just starting to see 400Gbps opportunities come across the desk. Fast forward to today and they seem to be almost as common as 100G requirements. I'm wondering what factors are driving this adoption the most as bandwidth requirements haven't grown 4x in a year. Crossconnect charges makeup a material amount of a circuit cost so above 100G it helps, cost per bit is less but hardware is potentially more, at least initially until fully used.

129

6 Comments