Anil Meena’s Post

💡 Can we Re-think AI-empowered Software Development? 💡 In 1895, Uriah Smith sketched a model for an automobile with a horse-shaped front. 🐴 This might seem odd to us now, but during his time it was the transition period from horse-drawn wagons to engine-powered cars. Smith was just trying to solve a relevant user problem — both horse wagons and cars were on the road, and the horses got scared of the cars. However, Smith was not capturing the true revolution of the engine. 🚘 He thought the major change was the replacement of one mobility machine — the horse — with another — the car. But the actual change was not about the type of machine; it was about the way we perceive mobility. Same goes for AI. 🤖 Tech leaders should spend some time imagining the future. Nobody knows exactly how AI-empowered software development process will look, but it will be certainly inherently different. You should shift from asking questions like, “Which AI tool helps my developers code faster?” to “How will be perceive the process of software development in the future?” 💬 Read the full article on Medium >>>

"Build-Measure-Learn" [1] and doing it quickly is a key idea popularized for a lean by Eric Ries. While this advice is focused on software startups, it also applies to hardware. With hardware this is easier said than done. The hardware iteration loop depends on many things including * #design time; * part procurement time; * build time; * #test time; and * #debug time. Of these, debug time can be the hardest to accurately predict and control. Debugging a mechatronic system [2] is both hard and necessary. Random variations cause embedded software to encounter things that were never expected when algorithms were designed and code written. Even with good #simulations and field data, it hard to get everything encountered at system integration [3], and afterwards, right. The behavior of the system changes in many unexpected ways due things like * manufacturing variations; * contamination from moisture and dirt; * corrosion; * electrial noise; * assembly errors; * damage; and * wear. These all combine to create undesired behaviors that are hard to explain and fix. Worse, with hardware its easy to end up with something that is tough to debug. Critical variables can be neither measurable nor practically observable [4]. This easily happens with physical quantities like a stress, pressure, temperature, or chemical composition in something like a car engine. It also happens with embedded controllers. At first it sounds strange that data cannot be observable in a piece of software. But with embedded controllers, Parkinson's Law [5] easily happens. The functions and code in the controller expand to use all of the available memory and throughput. When this happens and a new problem arises, there is not always an easy way to get data out and debug a problem. An inability to debug a system can kill a schedule as a team struggles to find a problem's root cause and fix it. Alternatively, a great team designs for and is able to debug the unexpected when it happens. But, how this is done is not always obvious. 😉 ------------------------------ For example, here is one #hack that works on low-cost controllers. This is handy when preferred methods of collecting data like screens and serial data messages are unavailable. Fortunately, with most #microcontroller applications, there are often unused pins. Also, with many modern microcontrollers, there are usually a timers available to drive a PWM [6] signal with almost no impact on real-time software operation. This gives an inexpensive way to get data out. A variable can be mapped to the PWM so it observable at all times. Then this can be measured with lab equipment like an oscilloscope to see what is happening. Even analog meters can be an ad-hoc monitor. This is simple, and when needed, works wonders. If you have a favorite debug hack, please share. #deeptech #hardtech Bonus points if you can guess what program is running on the Arduino below and what the PWM is measuring!

the best way to use whatever gp4 or the you know latest Clos Source models that are most expensive and have the most parameters uh just think of it as a prototyping tool anything you do with those prompts you can get your own model to do with a little bit more training it's kind of like uh when people build Hardware you have the analogy of uh prototyping with fpga which are very expensive right and then when you have the right architecture for Hardware then you do the circuit path and actually do the custom s so so right now for some of these tasks the large language model is sort of like your fpga whatever GPT 4 and then when you customize it you do like the super efficient one coding path for I don't know Shopify for coding assistance and Hardware software Etc that becomes your so that you train and customize which is cool I think that patterns emerging it's like as I hear you talk about that what's I just think it's just like so many different startups that could be built it just feels like we've never had this moment at least I didn't feel like I've never experienced a moment where there's just so many potential startup

Chiplets have gained traction recently in the news, with startup Baya Systems receiving $36 million in a series B round of funding. Arm recently announced progress with their chiplet system architecture (CSA), with 60 companies engaged in the technology. On Tuesday Keysight announced the launch of their Chiplet PHY Designer software. Fidelis Companies

Dive into the future of hardware engineering with our latest blog post by Feyza Haskaraman! Explore how technology evolution is reshaping innovation and uncover the potential for new ingredient technologies to revolutionize the industry https://lnkd.in/gri2hByp

Believe it or not, but AutoTech Detroit is letting me speak. 😃 If you're there, come fight me on this: If data is the source code of ML, then we need to iterate on it - after all, you would not attempt to build complex software by specifying 20 million lines of code upfront. Like a codebase, such datasets are necessarily an object under development. However, data sourcing reality all too often plays out differently: Some RfQ defines data needs 2 - 3 years into the future, before anyone can know how effectively those specs will improve model performance. The resulting high-volume, long-term contract is a good way to save a few cents per label - while wasting big bucks on the TCO of your development effort. So next week in Novi, I will pick a fight. A fight with the legacy principle of component cost optimization, applied to ML development and data sourcing. If your teams experience that challenge: Stop by the Suburban Collection Showplace on June 6 and let's talk about better ways to do product-grade ML.

🚀 Exciting News for Startups!🚀 Pricing and funding can be particularly challenging for startups, especially in our region. That's why I'm thrilled to announce that we offer specialized startup packages designed to lower the barrier to entry for software solutions at Startup Tailored Pricing. Our Mission ? To empower startups, fuel innovation, and accelerate your pace towards success.  #Startups #Innovation #SoftwareSolutions #Entrepreneurship #TechStartups #AI #Robotics #Automotive #Finance

Thanks to EE Times for their continuous coverage of Eliyan and one our key value adds to facilitate chiplet-based systems implementation "Chiplet technology is not new, but it is only in the last few years that uptake has driven the need for best practices, standards and tools to implement them. In late 2022, Silicon Valley startup Eliyan Corporation came out of stealth mode offering a more efficient approach to packaging. Its “bunch of wires” (BoW) chiplet system aims to achieve similar bandwidth, power efficiency and latency as die-to-die implementations using advanced packaging technologies by using standard packaging."

Technology is not just about code; it's about creating solutions that change lives. Every line of code is a step closer to shaping the future.

Outdated tools have stifled progress for hardware engineers. Reliance on PDFs, spreadsheets, and email makes it hard for product designers, supply chain managers and manufacturers to collaborate effectively. But all that is changing. The latest hardware engineering and manufacturing tech market map from Menlo Ventures shows the hot solutions for Design, Data Management, Manufacturing and Quality. And Duro's proud to be included. Here are the top trends to keep an eye on as the hardware engineering and manufacturing revolution takes off: 💡Removing collaboration barriers: Enabling seamless teamwork across design, engineering, and manufacturing, with out-of-the-box integrations between tools. 💡Harnessing AI for faster, smarter development: Automating tasks, generating design variations, and replacing time-consuming simulations. 💡Cloud-powered efficiency: Unlocking remote access to designs, centralized data and streamlined workflows across the entire hardware development process. https://lnkd.in/gNA-PqFi #hardwareengineering #hardwaredevelopment #manufacturing #agileworkflows