Anmol Darak

Heading to PyTorch India (Bangalore) — excited about vLLM × NVIDIA inference Large-scale LLM inference has quietly become a systems problem, not a model problem. As PyTorch India approaches, I’m particularly excited about discussions around: • vLLM internals (PagedAttention, continuous batching) • GPU-aware inference scheduling • Memory bandwidth + KV-cache pressure • Where PyTorch abstractions end — and hardware realities begin With sponsors like NVIDIA, IBM, and Red Hat, this event sits right at the intersection of: PyTorch → vLLM → GPU kernels → production reliability. Questions I’m carrying into the event: 1.What breaks first in real production — scheduler, KV cache, or memory bandwidth? 2.Where do PyTorch users *overestimate* batching gains? 3. How far can GPU-native optimizations take us before architectural changes are unavoidable? Looking forward to learning, debating, and connecting with folks building inference at scale.

18

Edge AI is gaining serious momentum 🚀 — and for good reason. Deploying models locally for data processing represents a meaningful shift in AI architecture — especially in regions where connectivity is inconsistent and data privacy concerns are rising, including across large parts of India. By moving inference closer to users, Edge AI can: Reduce latency and enable real-time performance ⚡ Strengthen data privacy by minimizing cloud dependency 🔒 Support functionality in low- or no-connectivity environments 📡 Lower cloud infrastructure costs for AI-driven products 💰 This shift can significantly expand access to AI-powered software for populations that were previously underserved 🌍. For startups, this is a strategic opportunity. With cloud costs being a long-standing pain point, optimizing for on-device deployment allows teams to focus more on product innovation and sustainable scale. While challenges remain — model optimization, hardware constraints, and deployment complexity — these are engineering problems that can be solved with the right architecture. Edge AI isn’t just a trend; it’s a structural shift in how intelligent systems will be built and deployed. Curious to hear your thoughts — is Edge AI becoming the default architecture for next-generation AI products?

25

1 Comment

𝐙𝐞𝐫𝐨 𝐭𝐚𝐱𝐞𝐬 𝐮𝐧𝐭𝐢𝐥 𝟐𝟎𝟒𝟕. Yes, you read that right. India just offered one of the boldest incentives in tech policy history and it’s aimed straight at the global AI and cloud industry. 👇 — In the 2026-27 Union Budget, India announced: 🇮🇳 Zero income tax on export revenue from cloud and AI services 🏢 Only condition? The workloads must run from data centres based in India 📅 Incentive lasts until 2047 - that’s 22 years of tax-free growth This is massive. For global cloud providers, it means they can serve the world tax-free if they plug into India’s infrastructure. For India, it’s a clear signal: “𝙒𝙚 𝙬𝙖𝙣𝙩 𝙩𝙤 𝙗𝙚 𝙩𝙝𝙚 𝙘𝙤𝙢𝙥𝙪𝙩𝙚 𝙘𝙖𝙥𝙞𝙩𝙖𝙡 𝙤𝙛 𝙩𝙝𝙚 𝙬𝙤𝙧𝙡𝙙.” — What’s included? ✅ Foreign cloud companies pay no income tax on revenue from non-Indian customers, as long as the workloads are processed in India ✅ Sales to Indian customers are taxed normally through local resellers ✅ 15% safe harbour cost-plus regime introduced for better transfer pricing clarity ✅ Huge global interest already: • Google: $15B in Andhra Pradesh • Microsoft: $17.5B by 2029 • Amazon: “Tens of billions” in pipeline • Adani & Reliance also ramping up — And here’s where it gets interesting: This isn’t just a tax play. It’s part of a larger bet - semiconductors, electronics, energy, and more. India is playing long-term infrastructure chess, not just AI checkers. But there are still questions: ⚡ Can India ensure stable power and cooling for dense compute? 🌊 Will water and sustainability challenges slow things down? ⏳ Can projects scale fast enough to make the most of this 22-year runway? — This changes everything. If executed well, India could become the go-to backend for global AI workloads. ♻️ Repost to share this shift with your network. #LeadWithAI #BuildWithAI #AIKnowledge #AIForBusiness #AILeadership #sushtend #AI #GenerativeAI #LeadersOfAI #LLM #GenAI

31

1 Comment

Role: AI Engineer CTC: 40+ LPA Question #12 / 24 ✅ In continuation of the 24-Part Series: DS/ML/AI Interview Questions I faced this year. Question: Your production model's performance has degraded over time. Why might immediately retraining the model be a bad idea? What would you investigate before deciding to retrain? CONCEPT: Imagine you trained a model 6 months ago. Initially -> Accuracy = 92% Today -> Accuracy = 81% The natural reaction is: Performance dropped → Retrain on newer data. But retraining assumes that the problem is caused by the model itself. What if the real issue lies somewhere else? Possible Causes: 1️⃣ Data Drift The incoming data distribution has changed. Example: The model was trained on customers from Region A. Now, most customers are coming from Region B. The model is seeing data it wasn't originally trained on. 2️⃣ Concept Drift The relationship between inputs and outputs has changed. Example: Customer behavior before and after a major economic event. The same features may no longer predict the same outcome. 3️⃣ Data Quality Issues A pipeline broke. A feature stopped populating correctly. Missing values increased. Retraining on bad data could actually make things worse. 4️⃣ Data Leakage Removal Sometimes a model performs worse because a previously leaky feature has been removed or corrected. The model appears worse, but is actually becoming more realistic. Retraining doesn't solve the underlying issue. 5️⃣ Label Issues The target labels themselves may have changed. If labels are delayed, incorrect, or inconsistently generated, retraining may simply learn from noisy labels. 🎯 Key Insight: Before deciding on retraining, we should first identify why the degradation occurred. Retraining can sometimes hide the problem instead of solving it. A useful framework: Performance Drop -> Investigate Drift -> Investigate Data Quality -> Investigate Labels -> Identify Root Cause -> Then Decide Whether Retraining Is Needed 💡 Interview Insight: This was a typical ML model-in-production question, with different possibilities for it failing. Please think more about production and deployment rather than just training and testing the model. Question #12 / 24 ✅ Next Question: AI Engineer role...Tomorrow, same time #AIEngineer #MachineLearning #MLOps #DataScience #ModelMonitoring #LLM #InterviewPreparation

80

2 Comments

Build Self-Learning Agents — No Fine-Tuning Required Fine-tuning isn’t always the answer. In our latest newsletter, Hari Ohm Prasath Rajagopal breaks down how you can build self-improving LLM agents without touching model training — using two simple but powerful patterns: 🔹 Dynamic Tooling: Let agents create and load new tools on the fly when existing ones aren’t enough. 🔹 Dynamic System Prompts: Allow prompts to evolve automatically based on real user or reviewer feedback. The key idea: Instead of expensive fine-tuning and slow human feedback loops, let agents learn from usage by evolving their tools and instructions in real time. He also walks through how this works in practice using Strands (AWS’s agentic framework), with minimal code and tight AWS integration. If you’re a software engineer building agentic systems and want: - Better domain accuracy - Faster iteration - Less ML complexity This one’s for you. 👉 Read the full article here: https://lnkd.in/dW3jtJ2W

16

1 Comment

A huge shoutout to @Strategink for orchestrating such a stellar #AgenticON2026. It was an incredible gathering of industry leaders from across sectors, all diving deep into the shift from traditional AI to truly autonomous systems. During my short session, I sat down to discuss a critical hurdle: Governance in the age of Agentic AI. In the traditional IT world, governance is static. We know the user, we know the API, and we set the perimeter. But as we move into a self-learning network, things change. AI agents are becoming Non-Human Identities (NHIs) with the power to act, reflect, and hand off tasks to other agents. Two key takeaways from the discussion: Identity is Dynamic: We can’t just authorize an agent by name. We must authorize it based on the specific task it is executing in real-time. Provenance is a Web, not a Line: Tracing an action is no longer about a single log file; it’s about auditing a hand-off chain across multiple tools and agents. The governance framework needs to address this chain of agency. Governance can’t be an afterthought; it has to be baked into the architecture. To my peers in the space: How are you addressing the identity crisis of AI agents in your own stacks? Are we ready for task-based authorization, or are we still relying on legacy models? What governance model or framework is your organization following? Let’s get a discussion going in the comments! 👇 #AgenticAI #AIGovernance #CyberSecurity #AgenticON2026 #FutureOfWork EC-Council DemandScience StrategINK Karthik S. Dr. Sayed Peerzade Stephanie Hills, Ph.D.

100

8 Comments

Wrote a blog explaining "LoRA Fine-tuning Efficiency Under Different Loss Functions." LoRA made fine-tuning cheap, but not always efficient. This post explores how loss functions like Cross-Entropy, Label Smoothing, and Focal Loss change convergence, stability, and generalization in lightweight LLMs. Read here: https://lnkd.in/gh_vm2kG

2