Dhamu Packiriswamy

🔍 Function Calls vs Model Context Protocol (MCP) in Agentic AI 🤖 Function calling brought structure to LLM interactions — enabling models to trigger tools, retrieve data, and execute tasks. But in agentic systems, where LLMs act as thinking agents rather than reactive bots, function calls alone don’t cut it. Enter: Model Context Protocol (MCP) 🧠 🔹 Function Calls = “Here’s a tool. Use it.” 🔹 MCP = “Here’s context, memory, history, toolset, persona, and environment — go accomplish a goal.” Why MCP matters: • ✅ Context-rich reasoning — gives the model structured access to user goals, memory, history, and tool capabilities. • 🔄 Multi-turn continuity — allows agents to persist state and adapt plans over long tasks. • 🛠️ Tool orchestration — lets the model choose and combine tools intelligently rather than just calling one-off APIs. • 📦 Modularity & Portability — MCP creates a standardized, modular interface for managing agent inputs and outputs — beyond raw JSON and ad hoc hacks. Think of function calls like giving an intern a list of API docs. Think of MCP like onboarding a smart assistant with access to calendars, tasks, Slack, past conversations, and team goals — and letting them figure out what to do next. Agentic AI needs structure beyond prompts. MCP is that structure — enabling autonomy, reasoning, and reusable frameworks at scale. We’re moving from “chat with AI” to “collaborate with AI agents.” #AgenticAI #MCP #LLMs #AIInfrastructure #ToolUse #AIContext #AIArchitecture #GenerativeAI #FunctionCalling

38

2 Comments

Shweta Shah, Ph.D. has suggested a concept that she calls "Content Quality Management". Shweta and I have been writing together for a while, and I am really impressed by this idea, that I think addresses exactly what the data community needs to consider, going forward in an era so indisputably dominated by AI. Shweta is adding to an ecosystem of thinkers like Bill Inmon (with his absolutely wonderful concept of the "textual warehouse") and Jessica Talisman (coiner of the "ontology pipeline"), and a wealth of people in the records- and information management space like tech entrepreneur Rachael Greaves, and skilled practitioners like Michael Upton, Andy Potter (deeply involved in ISO standards on AI records), and Stephen Clarke MRIM that are trained in thinking in the intersection of language and data. If you want ideas in how to succeed with AI, I suggest you follow Shweta Shah, Ph.D. #contentquality #contentqualitymanagement #ai #data #language #dataquality #dataqualitymanagement

42

6 Comments

A Thought Exercise : Context Engineering with LLMs is analogous to Feature Engineering in Classical ML What are the similatieties and differences ? Share in the thread #AImusings courtesy: Jim Hong :)

28

9 Comments

Humbled to see my article on recent H1B issue published in Amar Ujala The piece reflects on how the recent hike in H1B visa fees is not merely about visas — it is a reminder for India to rethink its digital self-reliance. Instead of sending billions abroad on software licenses and staying overdependent on the U.S. market, we can: ✅ Channel government budgets into Indian IT & AI innovation ✅ Build a sovereign digital backbone ✅ Create opportunities where Indian companies lead, not follow The road to आत्मनिर्भर डिजिटल भारत is closer than we think, if we choose to walk it. Curious to hear your views: Can India turn this H-1B crisis into its biggest innovation opportunity? #DigitalIndia #Innovation #H1B #AatmanirbharBharat #Technology

49

3 Comments

🚨 A new Data Radicals episode is live — and this one’s a masterclass in data products for AI. I sat down with Sanjeev Mohan — former Gartner analyst, author of Data Products for Dummies, and host of the It Depends podcast — to unpack what data products really are, how they fuel trustworthy AI, and why AI agents will reshape the future of work. Here are 3 insights that hit home: 🔁 Data products provide a "contract" between data producers and consumers. This builds trust and reuse. With ownership, documentation, and discoverability baked in, data products break down silos and finally make data usable – for data consumers and AI models alike. 💡 Data products bring software thinking to data. “We always buy software with a version number… but when was the last time we ever talked about data having a version number?” Data products shift the focus from raw artifacts to trusted, managed assets. 🤖 Your AI projects may be failing because they lack a data-product foundation. Running AI agents atop trusted data products boosts reliability and mitigates hallucination risk. As Sanjeev points out, “Data products give us the trust layer. What if I put my agents, my assistant, my chatbot on top of a data product?” Want to innovate with AI? You’ve got to experiment. As Sanjeev put it: “Unless you experiment, how do you know?” 🎧 Tune in and learn why Sanjeev calls this "the golden age of data": 👉 https://lnkd.in/gw7655GB #DataRadicals #AI #DataProducts #AIagents #DataGovernance #Metadata #FutureOfWork #Podcast #LLMs #DataLeadership

64

4 Comments

The news of a $100,000 H-1B visa renewal fee feels incredibly harsh for Indian professionals in the U.S. At first, it looks like another “shock” tactic — something extreme that eventually gets bargained down. We’ve seen this before: restrictions on AI chip sales to China that later turned into a 15% tax. But even as a proposal, this sends ripples across the industry: 🌍 Short term: Companies may accelerate offshoring. That could open more opportunities for freshers in India, and for students in U.S. universities as firms look for alternatives to costly renewals. 💻 Reality check: A large share of the U.S. tech workforce is Indian or Chinese. At places like Amazon and Uber, I noticed a pattern — the tougher the environment, the more you see immigrants. Not because it’s easier for them, but because survival and resilience are part of their journey. ⏳ Long term: America once proudly said “immigrants built this country.” Now the narrative feels closer to “protect local jobs.” If that mindset hardens, the U.S. risks losing its edge as top talent may simply choose Canada, Europe, or Asia instead. This also reminded me of something from my civics classes: India’s founding leaders debated whether to adopt a presidential system. The planning committee (led by Ambedkar and others) felt too much power in one individual could be dangerous. As a student, I believed a presidential system might have enabled faster decisions. Today, I see how that foresight may have guarded us from concentrated power — something we see tested in democracies worldwide. Whether it’s global tech policy or political systems, checks and balances matter more than speed. 👉 What do you think — will this visa fee push more roles offshore, or are we witnessing a bigger reset in how global talent and power structures are managed? My own perspective has shifted. I once believed the U.S. would always welcome immigrant talent. Today, I’m not so sure. What do you think — will this just push more roles offshore, or are we watching a bigger reset in how global tech talent is sourced? #H1B #Immigration #FutureOfWork #USIndia #TechTalent

10

1 Comment

 🧫 “Every cancer treatment starts with a single slide — and one critical judgment call.” But what if that judgment could be made faster, and with far greater precision? DigiDxDoc is building the bridge between traditional pathology and precision medicine. Jidhu Mohan Vysakh Mohan M Their tech platform digitizes glass slides and layers AI on top — removing subjectivity and standardizing outcomes in a space where accuracy is survival. This shift isn’t just technological — it’s philosophical. Pathologists, once limited by physical scope and human interpretation, are now empowered by digital workflows and computational insights. The result? Quicker diagnostics, more consistent care, and a step closer to true precision medicine. For any founder navigating the chasm between domain expertise and digital disruption, this story is a must-watch. 🧠 Innovation in healthcare doesn’t just save time — it can save lives. 👇 What’s the one manual process in your industry that needs an AI-powered overhaul? TapstartX #Community #StartupInnovation #HealthTech #DigitalTransformation #AIinMedicine #FounderJourney #PrecisionDiagnostics

29

1 Comment

Human level intelligence (AI) researchers need a dose of ancient Indian wisdom on intellect. I keep hearing debates on how we’ll reach human-level intelligence. Panels, podcasts, research talks… you name it. One of them featured the brilliant researcher Elaya Sutskever, who went deep into how human intelligence differs from what today’s LLMs give us. And every time I hear these discussions, I feel everyone’s missing a rather large elephant in the room. A very old, very Indian elephant. Our Rishis ( ancient Indian sages and researchers) already mapped this territory thousands of years ago. 𝗜𝗻 𝗜𝗻𝗱𝗶𝗮𝗻 𝘁𝗵𝗼𝘂𝗴𝗵𝘁, 𝘁𝗵𝗲 𝗶𝗻𝘁𝗲𝗹𝗹𝗲𝗰𝘁 𝗱𝗼𝗲𝘀𝗻’𝘁 𝗵𝗮𝘃𝗲 𝗼𝗻𝗲 𝗼𝗿 𝘁𝘄𝗼 𝗽𝗮𝗿𝘁𝘀. 𝗡𝗼𝘁 𝗲𝘃𝗲𝗻 𝘁𝗵𝗿𝗲𝗲. 𝗜𝘁 𝗵𝗮𝘀 𝘀𝗶𝘅𝘁𝗲𝗲𝗻 𝗮𝘀𝗽𝗲𝗰𝘁𝘀. Sixteen. Most of us can barely manage two on a Monday morning. Because sixteen is a lot for the human head (mine refuses after four), the tradition simplifies them into four broader categories:  • Manas  • Buddhi  • Ahankar  • Chitta If you grew up in India, you don’t even need definitions for Buddhi, Ahankar and Chitta. We’ve all used these words, usually during IPL matches or family arguments. Manas is the only one we quietly Google. Here’s the funny part. Almost all AI research today focuses on just one of these: Buddhi. Pure reasoning. Great for exams. Not great for building something that resembles holistic intelligence. And then we sit around wondering why AGI still looks like that relative who says “I’ll come by 6” and shows up at 10. With snacks. If ancient thinkers identified sixteen aspects of intellect, and we are barely scratching one… well, the math doesn’t look promising. It’s like trying to recreate a full South Indian thali with only sambar. You can try, but good luck explaining it to the waiter. My take? If we ever want machines to truly match human intelligence, we can’t keep polishing just Buddhi. We’ll have to tackle all aspects — yes, all sixteen — no matter how inconvenient that number is for modern PowerPoint slides. Until then, AGI will remain like my gym membership: theoretically possible, practically distant, and occasionally mocked. What do you think — are we overestimating Buddhi and underestimating the rest?

138

2 Comments

But yes… keep using ChatGPT for poems and Gemini for weekend prompt experiments. Meanwhile, at a national scale, AI analyzed 60 TB of data across ~177,000 restaurants ... impacting a country of 1.4+ billion people (read India) and helped uncover $8.4 billion in suppressed sales in India’s food sector. This is the “boring” side of AI: • Pattern detection across massive financial data • Anomaly tracking across states • Connecting dots humans would never see Just a small $8.4 billion issue affecting one of the world’s largest populations. No filters. No trending prompts. AI isn’t a toy anymore ... It’s infrastructure... It’s enforcement... It’s serious business. But yes… tell me again how your AI caricature got 300 likes on LinkedIn :)

35

1 Comment

🌍 SVO vs. SOV — Why Word Order Matters in AI Most LLMs like GPT-4 are trained on SVO languages like English. But languages like Hindi, Tamil, Marathi, and Bengali follow SOV structure — where the verb comes last. 🔹 English (SVO): “I eat rice.” 🔹 Hindi (SOV): “मैं चावल खाता हूँ” (I rice eat) This isn’t just grammar trivia. It’s a fundamental challenge in AI. LLMs trained predominantly on English struggle to understand or generate SOV languages fluently unless fine-tuned properly — or built from the ground up on Indic data. This is why building truly multilingual LLMs isn’t just about translation — it’s about respecting the cognitive architecture of each language. India’s linguistic diversity isn’t a bug — it’s a feature. Let’s build AI that reflects that. #AI #LLM #NLP #IndicLanguages #MultilingualAI #GPT #OpenSource #LanguageTechnology #LLMinIndia #SVOvsSOV #BhashaAI

472

22 Comments