Vimal Kumar Selvam

Doing more with less is what I have gotten into these days. I have seen BOM and demands coming to me/my team in TBs of storage and GBs of Memory and nearly half a dozen cores of 1-2Ghz. Today I am developing software for systems with less than 4MB of storage, 320KB of memory and two cores giving me 80Mz. Embedded systems and IOT were part of a magical world for me until few months ago. Simple things like back light of an LCD screen on a panel going off after few seconds, clock ticking away on the microwave, the beep when I turn on/off the fan or aircon, all were expected behaviour. Now that I am doing it myself, I am having the Chuck Nolan moment of having made his fire at that island. Coming from a time when 16Mz PC-AT using 640kb of memory and 1.44 even 360kb floppy disks was luxury, I see my current hardware is way superior to in early 90s. I used to juggle memory then through overlay files and TSRs, creating linked lists as and when I see need to store/purge data. INT21H was my thing for event driven coding. HiMem the one beyond 640kb I used it as RAMDisk which is like blazing fast drive. Brings back memories. The world of embedded systems hardware and the ecosystem has matured to a point where I am seeing otherwise heavy HTTPS, gRPC and Node.js implementation seen possible in such tiny bit of hardware. Host to Host API calls, is just few lines of code away. The capability of a platform like Espressif which gives BLE, WiFi with station and AP mode to coexist, it's ability to interface a ton of peripherals it's amazing. These have been around for years, I at times feel - where was I? Who knows, or one day AI systems which rely on blazing fast GPU hardware may even find its way into these tiny nano scale hardware. May be it already has also. #embeddedsystems #iot #domorewithless

10

This is the response to people who still ask questions on scalability and performance of blockchain technology. Peformance is depends on various factors like network topology, chaincode language, SDK, Payload, deployment etc. Hyperledger Fabric deployment in the cloud with limited compute achieved 2500 TPS for invoke(write transactions) and 3762 for query transactions(read). #blockchain #hyperledger #fabric #scalability #performance Pavan Adhav Hyperledger Foundation LF Decentralized Trust India Chapter Hyperledger Fabric

15

2 Comments

Good chatting with you Nitin Dahad at EE Times | Electronic Engineering Times on the emerging importance of #AI #Agents for #chipdesign. Watch the podcast at https://lnkd.in/gYiaarCH. Read my recent blog at https://lnkd.in/g4GvZFap.

51

1 Comment

📶 𝐏𝐢𝐨𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐭𝐡𝐞 𝐄𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧 𝐨𝐟 𝐖𝐢-𝐅𝐢 𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 At the forefront of advanced R&D, standardization, and successful commercialization, our journey through Wi-Fi innovations has been transformative. Here's a look at some key milestones: 🚀 802.11𝐠 - Drove 𝐎𝐅𝐃𝐌 𝐢𝐧𝐭𝐨 𝐭𝐡𝐞 802.11𝐠 standard. - Implemented OFDM in the first commercial CMOS 5GHz product, rapidly ported to 2.4 GHz. - Enabled the successful proliferation of 802.11g, greatly expanding the Wi-Fi ecosystem. 🏢 𝐖𝐢-𝐅𝐢 𝐌𝐞𝐬𝐡 - Pioneered 𝐦𝐮𝐥𝐭𝐢-𝐚𝐜𝐜𝐞𝐬𝐬 𝐩𝐨𝐢𝐧𝐭 (𝐀𝐏) technology. - Led contributions to the Wi-Fi EasyMesh industry specification. 📡 𝐖𝐢-𝐅𝐢 5: 802.11𝐚𝐜 - Drove 𝐌𝐔-𝐌𝐈𝐌𝐎, the foundational technology, into the 802.11ac standard. - Proved the viability and value of MU-MIMO for Wi-Fi systems. 📶 𝐖𝐢-𝐅𝐢 6: 802.11𝐚𝐱 - Drove key 𝐔𝐋 𝐎𝐅𝐃𝐌𝐀 𝐚𝐧𝐝 𝐔𝐋 𝐌𝐔-𝐌𝐈𝐌𝐎 𝐝𝐞𝐬𝐢𝐠𝐧𝐬 into the 802.11ax standards. - Developed synchronized AP scheduling, trigger-based OFDMA, and MU-MIMO. 💡 802.11𝐚𝐝/𝐚𝐲 - Pioneered 𝐦𝐦𝐖𝐚𝐯𝐞 𝐛𝐞𝐚𝐦 𝐟𝐨𝐫𝐦𝐢𝐧𝐠 & 𝐚𝐧𝐭𝐞𝐧𝐧𝐚 𝐝𝐞𝐬𝐢𝐠𝐧𝐬. - Delivered the highest performing mmWave implementation, proving its viability. - Brought wider bandwidths via channel bonding in 802.11ay. 🌐 𝐖𝐢-𝐅𝐢 4: 802.11𝐧 - Implemented 𝐌𝐈𝐌𝐎-𝐎𝐅𝐃𝐌 system architecture for mass market products. - Drove the success of pre-standard MIMO products. 📡 802.11𝐚𝐡 - Enabled extended range, 𝐬𝐦𝐚𝐥𝐥 𝐛𝐚𝐭𝐭𝐞𝐫𝐲 𝐨𝐩𝐞𝐫𝐚𝐭𝐢𝐨𝐧, 𝐚𝐧𝐝 𝐥𝐨𝐰-𝐩𝐨𝐰𝐞𝐫 for Wi-Fi below 1 GHz. - Facilitated the opening of spectrum for 802.11ah sensors in Europe. 🔜 𝐅𝐮𝐭𝐮𝐫𝐞 𝐈𝐧𝐧𝐨𝐯𝐚𝐭𝐢𝐨𝐧𝐬 - Continued leadership in IEEE and WFA. - Ongoing work on 𝐖𝐢-𝐅𝐢 7 / 802.11𝐛𝐞. Our commitment to innovation in Wi-Fi technology has paved the way for faster, more reliable, and more efficient wireless communication. Here's to many more breakthroughs in the future! Source : IEEE Qualcomm #WiFi #Innovation #Technology #Wireless #Networking #Standardization #MU_MIMO #IEEE #WFA #WiFi6 #WiFi5 #4G #5G #NetworkArchitecture #EdgeComputing #IoT #NetworkSlicing #LowLatency #HighThroughput #MassiveMIMO #Beamforming #Virtualization #ML #AI #Telecom #Innovation #Technology #FutureOfConnectivity

5

An IEEE SA whitepaper authored by ADWG members have been published. -Simulation, Testing, Verification, and Validation (STV2) of Autonomous Driving. It can be downloaded at https://lnkd.in/gvVsF5P3.

10

Satoshi -San did an amazing job of highlighting Quest Global'a differentiating potential in the semiconductors space at his talk. Global collaboration is made possible by a focus on client relations and delivering quality results! #Semiconductors #Leadership #BusinessStrategy

5

“Make in India” strategy now adds “Make Silicon in India” - Prime Minister Narendra Modi said India’s time to lead the 'Silicon Diplomacy' has arrived. - “When the chips are down, you can bet on India… India is set to play a big role in driving the global semiconductor industry,” - PM Modi said the government has made special efforts to ensure it provides not just financial incentives (with central govt offer 50% financial support for setting up semiconductor manufacturing facilities) to companies investing in semiconductors, but also gives them ease of doing business. - With India’s electronics sector touching over $150 billion in revenue, the PM outlined a larger goal to grow this to $500 billion by 2030 - Actions are already taking place, Tata group to build First mega silicon manufacturing facility in Dholera, Gujarat in partnership with Taiwan’s Powerchip Semiconductor Manufacturing Corporation (PSMC) with Indian engineers are already undergoing training by PSMC All the best #GOI for the full swung initiative , looking forward to see India’s contribution to Global semiconductor demands.

16

Thanks People+ai! Check out the video below to know what I am up to these days! I am building hardware to make AI accessible to all!

59

5 Comments

Excited to be presenting at The Future Memory and Storage 2024 (FMS'24) next week, with Navneet Rao. As data sets get larger, optimizing data movement between the different compute elements CPU, DPU, GPU etc is critical to achieving better efficiency and TCO. This talk focuses on how CXL can be used to minimize data movement when a processing pipeline is composed of heterogenous compute elements. Thanks Ajay Gupta for the support. Paper Title: Improving SW-HW processing pipeline for storage stack/service workflows w CXL Paper Abstract: Storage services & workflows are adopting disaggregated storage architecture enabled by newer ethernet standards: 400/800GE. This also necessitates need for HW accelerators for operations like compression, encryption, erasure coding, deduplication etc with some usages requiring operations at storage initiator. These operations require data context to be translated from memory into storage context right at the Storage initiator (typically, CPU) before transmission over 400/800GE to the destination media block/object/file storage. The storage stack thus is a blend of protocol processing & multiple translations enforcing data segmentation using SW-based execution or through accelerator usage (e.g., TLS encrypt / decrypt, NVMe data digest, etc.). Almost always, requires multiple data movement between cpu memory domain and lookaside accelerator memory domains with PCIe link becoming the bottleneck. This paper proposes a CXL based implementation to preserve memory context of data as long as possible, chaining storage and network functions to deliver higher performance for both local & disaggregated storage accesses: block, object, file across varied storage media: volatile, ephemeral, non-volatile. The solution will enumerate CXL concepts to enhance current IO paths and show how adopting CXL dataflow (Type1 / Type2 / Type3) allows for data plane in only one memory domain with control plane straddling between cpu and accelerator memory domains. This delivers higher throughput, more IOPs per node & also reduces deployment cost by enabling solution upgrades requiring slower hardware refresh cadence.

65

2 Comments

I periodically read the Yugabyte blog, and just stumbled across this gem by Karthik Ranganathan. >Organizations can start for free with open source software without speaking to sales or hitting a demo request or contact us form. This makes exponential adoption growth a real possibility. I would agree—walking back freely-available software to be no longer free is not a good business plan. When a LittleHorse competitor ( Camunda ) did this, an advisor of mine suggested sending them a thank-you-note 😉 Long story short: if you are currently using Camunda or CockroachDB, check out LittleHorse and Yugabyte, respectively! https://lnkd.in/eUdqWmaY

35

2 Comments

So many wonderful memories from AMD's Advancing AI Day yesterday. If I had to pick one highlight, it would be our startup panel with Vamsi Boppana, our SVP of AI, alongside my favorite entrepreneurs. 1. Ashish Vaswani from Essential AI 2. Dani Yogatama from Reka AI 3. Amit Jain from Luma AI 4. Dmytro Dzhulgakov from Fireworks AI These founders have been pushing the boundaries of what’s possible with our MI300X GPUs, helping us confidently meet tomorrow’s AI demands today. Together with our ecosystem partners, they are a driving force behind advancing AI for everyone.

88

2 Comments

In my first job at Qualcomm, I was part of a ~30 person team where I was the only non Indian/Pakistani. And I loved it! You would think it would create a weird dynamic. That I would feel isolated, wouldn't fully integrate, and have little in common with my desi colleagues. But that couldn't be further from the truth. I was welcomed with open arms, and I fully embraced it 👇 We ate tons of Indian food and had chai breaks. (I loved chicken vindaloo & mango lassi) We watched Bollywood movies during bring-up days. (My favorite was Singham & Robo) I started inadvertently doing the head wobble. I learned all about cricket and we played TT. I was invited to weddings in India. I picked up some Hindi phrases. I was even referred to as Alex bhai! It often times felt like I was truly in India. So I look back very fondly on those first 7 years of my career. It was such a warm and welcoming environment for a fresh graduate. I made so many wonderful friends & memories. We could have easily chosen tribalism, and kept to our own "kind". But instead, we chose unity. We celebrated our differences, not used them as a way to create separation. In times like these, where it's so easy to get caught up in the "us vs. them" narrative, it's important to remember that there is much more that unites us than divides us. So tune out the noise which commands you to hate. Choose togetherness. Choose courage. Choose love 💙

91

15 Comments

🎉 While New Year’s celebrations are in full swing across ex-USSR territories, I took a different route this time—focusing on productivity! 🚀 I dove into some experiments with #Bazel and #Cplusplus, setting up a C++ infrastructure from scratch. Despite my extensive experience with Bazel, this was my first time building a C++ codebase with it from the ground up. What can I say? The Bazel experience is a breath of fresh air compared to the more widely used #CMake. Its simplicity, modularity, and efficiency make it a standout choice for managing complex projects. If you haven’t tried it yet, I highly recommend giving it a shot for your next C++ project. Here’s to starting the new year with fresh experiments and new insights! 🥂 Have you explored Bazel for C++ or other languages? I’d love to hear your experiences! https://lnkd.in/gsphC_62

9

The recording of my talk at Arm DevHub Live on 'Using Zephyr to Manage Hardware Supply Chains' just went live: https://lnkd.in/dyu2p8zu It is a short summary of our experiences being an OEM device maker over the last several years at mBedrock and why we rely on The Zephyr Project Does the premise seem far-fetched to you? If not, why aren't you using #ZephyrRTOS yet? 😀 #ZephyrRTOS #Linux #mBedrock #PCBKingdom

28

Counterpoint Research’s Ashwath Rao examined how the latest innovations in etch will allow chipmakers to scale to higher layers of 3D NAND, which is critical to unlocking future AI applications via his latest EE Times Asia article. Read it now to learn how Lam Research offers groundbreaking solutions to common scaling challenges. #LifeAtLam

10

The shift from #AI in the cloud to AI on-device is going to bring huge benefits in performance, personalization and privacy. Learn the basics of on-device AI in this upcoming DeepLearning.AI session, hosted by Qualcomm Senior Director of Engineering Krishna Sridhar #TeamQualcomm https://lnkd.in/gGc4Ef5E

69

Summarized versions of various project solved so far: 1. Semiconductor Manufacturing Yield Prediction: Utilized supervised learning, ensemble modeling, and unsupervised learning to build a prediction model identifying Pass/Fail yield in semiconductor manufacturing, aiming to enhance process throughput and reduce production costs by pinpointing key factors in yield excursions. 2. Vehicle Segmentation and Classification: Part 1: Segmented cars based on fuel consumption and other attributes. Part 2: Classified vehicle silhouettes into one of three types using extracted features, considering various viewing angles. 3. Telecommunication Customer Churn Prediction: Developed a machine learning workflow to predict customer churn for a telecom company, aiding in customer retention by analyzing patterns in data and running autonomously with CSV files to identify the best-performing model. 4. Healthcare and Marketing Prediction: Part 1: Predicted patient conditions based on biomechanics features and test results. Part 2: Built an AIML model for focused marketing, predicting potential customer conversions using historical data, after performing extensive EDA and addressing missing values and imbalanced data. 5. Statistical Analysis and Decision Making: Part 1: Solved industry problems using statistical inferences. Part 2: Evaluated the status of startups participating in a prominent startup competition, leveraging statistical evidence for effective decision-making.

15

Yocto plays a significant role in the deployment of our software stacks for end product use cases. There were couple of key presentations from my team at Texas Instruments at ongoing Yocto Project Summit : 1) Leveraging Dynamic Layers in Yocto Project by Chirag Shilwant https://lnkd.in/gjiRKHJY 2) Supporting Multiple layer setup tools by Ryan Eatmon https://lnkd.in/gPp4h_WU These are generic concepts that should help the Linux community using Yocto. The Linux Foundation , Linux, The Yocto Project, A Linux Foundation Project

91