Shashank Agrawal

**Embracing What We Can't Control** In life, there are two events we can't control: our birth and our death. These important moments are out of our hands, and this can help us see that worrying about other things we can't control doesn't make much sense. Instead, we should focus on giving our best effort and let the rest happen as it will. When we accept that we can't control everything, we can stop stressing over things that don't go our way. Instead of worrying, we should put our energy into doing our best in every situation. This way, even if things don't turn out as we hoped, we know we did our best. "Give your best and leave the rest" can guide us. It reminds us that while we can't control every outcome, we can control our actions and attitudes. By adopting this mindset, we can be more resilient, find peace, and enjoy the possibilities life offers without being overwhelmed by worry. Understanding that we can't control life's beginning and end helps us focus on living fully in the present. It teaches us to use our energy wisely, accept uncertainties, and trust the journey, knowing we've done our best. #LiveFully #LetGoOfWorry #DoYourBest #EmbraceLife #MindfulLiving #FocusOnNow #TrustTheJourney #LifePerspective #PSG

11

India's focus on achieving self-reliance in technology is both ambitious and crucial. While the journey is fraught with challenges—such as cultivating local talent, fostering innovation, and establishing a robust R&D ecosystem—the potential benefits are immense. Success in this endeavor will not only strengthen India's technological independence but also drive economic growth and enhance global competitiveness.

5

1 Comment

Everything gets RESET. Story of my friend Akash was very similar to IP. He changed his job and realised that it was like everything got reset for him. Watch at 1.5x speed the story of Akash in short video and then if interested to learn more, put the second video in your watch list. Both videos are in comment section. This will help you to understand one of the successful Campus Network solution. Old videos of mine but still worth to watch. Happy Learning !!!

19

4 Comments

As we approach New Year, people will start posting their grand plans for the next year. However, it is worth spending some time to retrospect - "how many goals you set for 2024 are executed till date?" Same is true in case of Corporates which boast fancy Vision Statements but fail to execute things at ground level. What are your thoughts on this? #mondaymotivation #vision #execution

12

Its true that though Indian Talent has a breadth of expertise to offer, we lack on in-depth knowledge (Theres a very small percentage as an exception to this) and one of the main reasons lies in our education system. For years, our education system incentivized "Timely completion of given task" and "memorizing information", but not quality and journey to completion. "How a particular task can be achieved in different ways and what shall be the best way" is not something many schools focus on. So, students lose the opportunity to spend more time on a particular topic and understand it better, because all they what to chase for are "Academic marks" Secondly, "Obedience to parents and teachers seen as a virtue" has killed student's ability to think openly on a topic and ask questions without being judged by parents and teachers. There are many other reasons , but if we at least address above two blockers, we can certainly cross many developed countries in Innovation in the long term . https://lnkd.in/geHR6XiQ

11

I usually avoid commenting on “people incidents,” but the Deepinder Goyal case raises fundamental questions on fairness, learning, relationships, charity. IMO, stripped of noise, the issue boils down to this: Would you pay ₹70 lakh (₹20 lakh "learning fee" + ₹50 lakh opportunity cost) to get a ₹50 lakh per annum job? Key Questions: Q: Aren't you being too transactional? What about the learning? A: Learning never stops. Even after 21 years, I’m still learning. Should I keep paying my employers? Q: Then when should a "learning opportunity" become a "paid job"? A: Ask: "If a learner refuses to pay, would you hire someone for a paid role to do the work?" If yes, the value of work exceeds the value of learning. MBA schools don’t hire people to solve case studies or attend lectures. Q: But the fee is being donated to charity. A: A donation is unconditional. If mandatory, it’s a fee — a deal. What happens to the money afterward is irrelevant. Q: But they’re also donating ₹50 lakh themselves. A: Their donation, whether 50 lakh or 5 crores is their choice. Such "match up" measures are used to encourage "voluntary donations" not "mandatory fees". Again, the difference between a donation and a deal is not the amount — it’s conditionality. Q: What is an "unconditional donation"? A: True donation comes from feeling prosperous: "I have enough and more, so I share." Most donations are deals — for branding, tax savings, washing sins, accumulating blessings. Very rich people can FEEL poor. People with less can feel prosperous. Q: That would make 99% of donations 'deals'? A: Yes. The problem isn’t making deals; it’s mislabeling them as donations, which kills the possibility of tasting true abundance. Consequences for Candidates: 1 - Leaving would be hard due to significant financial and emotional investment — like China’s “vendor financing” for the U.S. 2 - Risk of toxicity or exploitation masked as “Real life is hard.” Risk, not certainty. 3 - A ₹50 lakh job over 5 years(if you leave) will equate to ₹36 lakh annually. 4 - Lot of fame. And reputation risks: known as someone who "bought a job" and, if leaving, as someone who “couldn’t handle the pressure.” Consequences for the org: 1 - This role is for less experienced candidates. They will either have cushy familial finance cushions or might need loans. This creates either "flings" or "bondage" 2 - Media visibility cuts both ways: celebrated as a “disruptor” or criticized as an “exploiter.” 3 - It filters out non-serious candidates, ensuring “skin in the game.” 4 - Attrition risk drops—fear of loss deters exits. 5 - But better mechanisms likely exist to achieve 3 and 4 above 6 - Fear-based loyalty isn’t commitment. When fear subsides, so does loyalty. Important to notice in an important relationship. Of course, the novel idea will still invite many takers. But a popular choice and wise choice are not the same thing. I hope all parties make an informed choice based on consequences.

5

5 Comments

For decades, we have lived with the below saying. USA innovates, China replicate, EU regulates, India adopts or operates, It's about time we rewrite this narrative. India is ready to break free from the confines of this saying. I envision a future where India is a hub of innovation and replication. #newindia #economy #innovation

9

1 Comment

Ex: vector addition in both HIP & CUDA CUDA #include <iostream> __global__ void vectorAdd(const float* A, const float* B, float* C, int N) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < N) { C[i] = A[i] + B[i]; } } int main() { int N = 1024; size_t size = N * sizeof(float); // Allocate host memory float *h_A = (float*)malloc(size); float *h_B = (float*)malloc(size); float *h_C = (float*)malloc(size); // Initialize input arrays for (int i = 0; i < N; i++) { h_A[i] = static_cast<float>(i); h_B[i] = static_cast<float>(i); } // Allocate device memory float *d_A, *d_B, *d_C; cudaMalloc(&d_A, size); cudaMalloc(&d_B, size); cudaMalloc(&d_C, size); // Copy data from host to device cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, h_B, size, cudaMemcpyHostToDevice); // Launch kernel int threadsPerBlock = 256; int blocksPerGrid = (N + threadsPerBlock - 1) / threadsPerBlock; vectorAdd<<<blocksPerGrid, threadsPerBlock>>>(d_A, d_B, d_C, N); // Copy result back to host cudaMemcpy(h_C, d_C, size, cudaMemcpyDeviceToHost); // Free memory cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); free(h_A); free(h_B); free(h_C); return 0; } HIP #include <hip/hip_runtime.h> #include <iostream> __global__ void vectorAdd(const float* A, const float* B, float* C, int N) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < N) { C[i] = A[i] + B[i]; } } int main() { int N = 1024; size_t size = N * sizeof(float); // Allocate host memory float *h_A = (float*)malloc(size); float *h_B = (float*)malloc(size); float *h_C = (float*)malloc(size); // Initialize input arrays for (int i = 0; i < N; i++) { h_A[i] = static_cast<float>(i); h_B[i] = static_cast<float>(i); } // Allocate device memory float *d_A, *d_B, *d_C; hipMalloc(&d_A, size); hipMalloc(&d_B, size); hipMalloc(&d_C, size); // Copy data from host to device hipMemcpy(d_A, h_A, size, hipMemcpyHostToDevice); hipMemcpy(d_B, h_B, size, hipMemcpyHostToDevice); // Launch kernel int threadsPerBlock = 256; int blocksPerGrid = (N + threadsPerBlock - 1) / threadsPerBlock; hipLaunchKernelGGL(vectorAdd, blocksPerGrid, threadsPerBlock, 0, 0, d_A, d_B, d_C, N); // Copy result back to host hipMemcpy(h_C, d_C, size, hipMemcpyDeviceToHost); // Free memory hipFree(d_A); hipFree(d_B); hipFree(d_C); free(h_A); free(h_B); free(h_C); return 0; } Kernel Definition: vectorAdd Memory Management: Each example allocates memory on both the host & device, copies data between them, and frees allocated memory afterward. Kernel Launch: The syntax for launching the kernel differs slightly (<<<...>>> CUDA & hipLaunchKernelGGL HIP). Data Types: The data types and operations are identical

1

Network design in data centers is evolving to accommodate the increasing demands of AI workloads. The primary changes revolve around higher bandwidth, reduced latency, and increased data movement efficiency. Trends 1. AI-Specific Infrastructure: Data centers are increasingly being optimized specifically for AI workloads, which require large-scale parallel processing and real-time data analysis. Traditional networking infrastructure is evolving to handle this by integrating high-performance interconnects, such as NVIDIA’s NVLink and InfiniBand, to provide low-latency, high-bandwidth communication between AI accelerators like GPUs and between servers. 2. Disaggregation of Resources: AI workloads benefit from the disaggregation of compute, storage, and networking resources, allowing for flexible and efficient resource allocation. For example, composable infrastructure is gaining traction, where resources like GPUs, CPUs, and memory can be pooled and dynamically assigned to different AI tasks. This enables data centers to handle fluctuating AI demands more efficiently. 3. High-Bandwidth Memory and Networking: To keep up with the growing size and complexity of AI models, data centers are deploying high-bandwidth memory (HBM) and next-generation networking technologies like 100G/400G Ethernet and Quantum Ethernet fabrics. These technologies enable faster data transfer between AI accelerators, reducing bottlenecks and improving overall AI training and inference performance. 4. Edge AI and Distributed Networks: With the rise of edge computing, AI processing is increasingly being distributed across edge data centers closer to where the data is generated. This reduces latency for AI applications that require real-time decision-making, such as autonomous vehicles and IoT systems. In this distributed network design, data centers are interconnected via high-speed, low-latency links to ensure seamless communication between edge nodes and central AI systems. 5. AI-Oriented Cooling and Power Efficiency: AI hardware, especially GPUs and TPUs, generates a significant amount of heat and consumes large amounts of power. Data centers are evolving to support these requirements with advanced cooling solutions like liquid cooling and immersive cooling. New network architectures incorporate dnergy efficiancy. 6. SDN and AI-Driven Automation: Data centers are leveraging SDN to manage the complex networking demands of AI. SDN allows data centers to dynamically allocate network resources based on AI workload needs, improving efficiency. Moreover, AI is now being used to automate network management, enabling predictive maintenance, performance optimization, and better traffic management, especially in highly dynamic AI environments. In summary, data center network design is moving toward higher performance, flexibility, and scalability, with a strong focus on AI-optimized infrastructure, faster data movement, and energy-efficient designs for AI driven applications.

1

This report is aptly titled: India - A visible snapshot. I want you to consume all this data taking an analogy of a spectacle, you are wearing. And assume the two lenses of this spectacle being prescribed for myopia (farsightedness) and hyperopia (nearsightedness). 😄 Taking the anology one level further into a case of pricing (nearsightedness) and adoption (farsightedness). While (at first read) the potential of this country in terms of scale and opportunity seems limitless, the aspects of pricing and adoption are still complex, and unless taken seriously, wont yield as much (enticing) as this data and its numbers. If this was too abstract/cryptic, let me close the post with an example: Despite the data on the chart behind the Internet users, Mobile subscribers, smartphone users etc.. The ARPU of all Indian operators are below par and Sunil Mittal (Airtel) says, we have the world's lowest ARPU (average revenue per user) at Rs.300 (<4$). I hope you get the myopia (pricing) angle now 😁. And adoption is far more complex. Do you have a example for me? Leave a comment. Remember always that Data is a double edged sword. #I❤️Writing

10

1 Comment

The exponential growth of digital payments in India, increasing nearly 90-fold over the past 12 years, is a testament to our nation's digital transformation. This surge, while impressive, underscores the critical need for robust security measures to safeguard against digital #payment fraud. To sustain this growth, we must empower fintech companies providing last-mile services. These companies are crucial in bridging the #credit gap, particularly for #SMEs and #Nanopreneurs, thereby fueling entrepreneurial growth and job creation. I believe exempting #GST on services offered through Business Correspondent outlets is essential. This tax relief would lower operational costs and encourage greater outreach, fostering financial empowerment at the grassroots level. Furthermore, I see the recognition of Self-Regulatory Organizations (SROs) in the fintech sector as a positive step towards better regulatory alignment and communication. This move will undoubtedly contribute to a more robust and responsive financial ecosystem. As we approach the #UnionBudget 2024, I see immense potential for our #financial services sector to drive significant economic growth and financial inclusion. #SpiceMoney #EmergingIndia #NewIndia #RuralFintech Ramesh Venkataraman Venkatram Jayanthy Rashmi Aggarwal CA. SUNIL KAPOOR MRUTYUNJAY MAHAPATRA Ram Rastogi 🇮🇳 Rohit Ahuja Harsh Mittal sameer nagpal Usha Murali Kuldeep Pawar Atul Tiwari https://lnkd.in/dr6ezcD6

34

1 Comment

A few months back, a junior colleague with 5-6 years of experience asked me about tax deductions. His question was simple yet revealing: *'Sir, thoda hi toh kat raha hai, isme sikhne ki kya zarurat hai?'* It struck a chord with me because I had the same mindset early in my career. Back then, I didn’t think much about taxes, salary breakdowns, or financial planning. I believed it wasn't something to worry about as long as I was getting paid. But over time, I learned that ignoring these details can cost you—literally. Here’s what I explained to him and what I believe every professional, no matter their experience level, should understand: 1️⃣ Know Your Tax Structure - Don’t ignore tax deductions just because they seem small. Over a year, they add up.  - Understand tax slabs and learn how exemptions like HRA, LTA, and 80C deductions can reduce your liability. 2️⃣ Decode Your Salary (CTC ≠ In-Hand)  - CTC is not what you take home. Break it down into components like basic pay, PF contributions, allowances, and bonuses.  - Ask for clarity on one-time bonuses, variable pay, and ESOPs. 3️⃣ Plan Investments Early - Start with tax-saving options like ELSS, PPF, or NPS to reduce taxable income.  - Create an emergency fund and invest in a mix of mutual funds, FDs, and insurance for long-term security. 4️⃣ Key Questions to Ask Before Accepting an Offer - What is the exact in-hand salary after all deductions?  - Does the company provide perks like health insurance, performance bonuses, or ESOPs?  - What’s the notice period and severance policy? => Why This Matters Financial literacy isn’t optional—it’s a life skill. By understanding and planning your finances early, you’re not just saving money; you’re building a foundation for future financial freedom. I wish someone had taught me this at the start of my career—it would have saved me from many missteps. If you’re starting, take the time to understand your pay, taxes, and investments. Trust me, it’s one of the smartest investments you can make in yourself. What financial advice do you wish you had received earlier in your career? Let’s share and help others grow! 💬 #FinancialPlanning #SalaryTips #TaxSavings #CareerAdvice #Linkedin #Linkedinconnections #VikrantG Financial Times

7

1 Comment

With rising demand for GPUs, the infrastructure providers need a cloud alike layer for offering their AI compute resources in a "as a service" model. Checkout this blog and a demo to know what Aarna has to offer in this space!

22

1 Comment