Amar Jaiswal

Had the pleasure of hosting Himanshu Girdhar on my podcast recently. Himanshu is currently working as Director in Data Analytics at American Express, with close to a decade of experience in driving data-led decision making across product and marketing analytics. What makes his journey particularly inspiring is the strong academic foundation—an alumnus of Delhi School of Economics (Batch of 2017), AIR 2 in the DSE entrance, and having cracked multiple top exams including ISI, JNU, IGIDR, and SAU. In our conversation, we spoke about: • How to approach top economics entrance exams • Transitioning from core economics to analytics • What really matters in building a career in data-driven roles • The intersection of economics, data, and real-world business decisions A lot of valuable insights for students and young professionals who are trying to navigate this space. Looking forward to bringing more such meaningful conversations through Arthapoint. https://lnkd.in/gY2DDTGX

31

India’s AI summit is a reminder of a policy trap: 𝗬𝗼𝘂 𝗰𝗮𝗻’𝘁 𝗰𝗼𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝘆𝗼𝘂𝗿 𝘄𝗮𝘆 𝗶𝗻𝘁𝗼 𝗮𝗻 𝗔𝗜 𝗲𝗰𝗼𝘀𝘆𝘀𝘁𝗲𝗺. If policy wants real research traction and industry pull-through, the levers aren’t stages and speeches. They’re incentives and institutions: • 𝗦𝘁𝗮𝗯𝗹𝗲 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗳𝘂𝗻𝗱𝗶𝗻𝗴 (5–10 year horizons), not annual headline budgets • 𝗖𝗼𝗺𝗽𝘂𝘁𝗲 𝗮𝗰𝗰𝗲𝘀𝘀 𝗮𝘁 𝘀𝗰𝗮𝗹𝗲 for universities and startups (transparent, merit-based) • 𝗖𝗹𝗲𝗮𝗿 𝗱𝗮𝘁𝗮 𝗿𝘂𝗹𝗲𝘀 that enable lawful research collaboration (privacy-safe, predictable) • 𝗧𝗮𝗹𝗲𝗻𝘁 𝗳𝗹𝘆𝘄𝗵𝗲𝗲𝗹𝘀: easier cross-border hiring + incentives to retain top researchers • 𝗧𝗲𝗰𝗵 𝘁𝗿𝗮𝗻𝘀𝗳𝗲𝗿 𝘁𝗵𝗮𝘁 𝘄𝗼𝗿𝗸𝘀: simpler IP terms, faster spinouts, procurement that can actually buy from startups The harsh truth: if your policy output is mostly “AI ambition” messaging, you’ll get exactly that—messaging. The countries that win will make it boring: reduce friction for 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵, lower risk for 𝗰𝗮𝗽𝗶𝘁𝗮𝗹, and shorten the path from 𝗹𝗮𝗯 → 𝗽𝗿𝗼𝗱𝘂𝗰𝘁 → 𝗺𝗮𝗿𝗸𝗲𝘁. What’s the single highest-leverage policy move India should prioritize: 𝗰𝗼𝗺𝗽𝘂𝘁𝗲, 𝘁𝗮𝗹𝗲𝗻𝘁, 𝗼𝗿 𝘁𝗲𝗰𝗵 𝘁𝗿𝗮𝗻𝘀𝗳𝗲𝗿? #AIStrategy #AIResearch #InnovationPolicy #TechEcosystem #IndiaTech

6

Recently, I came across news about PhysicsWallah’s new AI model that beat GPT-4 on JEE math questions. Naturally, the internet split in two—some proudly calling it India’s moment, others waving the “it’s just fine-tuning” flag. Here’s the reality: This is a small language model, not a large one. And its job wasn’t to write novels or generate Python—it was built specifically to excel at JEE math. Which it did, remarkably. But here’s where many get it wrong. Just because a domain-specific model outperforms GPT-4 on one task doesn’t make it more “intelligent.” That’s like saying a calculator is better than a scientist because it computes faster. So I wrote this piece to help business leaders, investors, and AI enthusiasts avoid falling for misleading headlines like “We built a model better than GPT-4 at X.” It probably is. But only at X. The article breaks down: • The difference in training goals, data, and infrastructure • Why smaller models are cheaper, safer, and sometimes better—but not general • How to spot hype vs actual product strategy Give it a read below if you’ve ever caught yourself thinking “why not just build our own GPT?” https://lnkd.in/gT-_JWnZ #AI #LLM #SmallLanguageModels #GPT4 #EdTech #DeepTech #AIProduct #MachineLearning #DomainSpecificAI #VC #AIForBusiness #LinkedInAI #AIethics #AIgovernance #LLMvSLM

30

1 Comment

Consuma: The CIA for Consumer Intelligence Consuma has raised ₹12 crore to reshape consumer research with near-instant insights. By delivering actionable intelligence in 30 minutes — instead of months — and leveraging real-time data extraction, Consuma is winning major clients including Britannia, Godrej, and Pepsi. The company aims to redefine the speed of consumer decision-making. #Consuma #MarketIntelligence

4

At Chi Square 2026 – National Analytics Conclave, organized by DARVIX XLRI. Jamshedpur The panel theme was “AI Beyond the Hype Cycle: Value Creation in a Bubble Era.” Summary of my position: Most organizations do not struggle with generating insights. They struggle with translating those insights into clear decisions — and ensuring those decisions remain coherent as actions unfold across the system. Between insight and action lies: * Human judgment * Decision Trade-offs * Authority and accountability * Feedback and learning AI can surface patterns at scale. But value is realized only when decision rights are clear, boundaries are defined, and outcomes are continuously reviewed. In complex enterprises, local optimisation is easy. System coherence is hard. Grateful for the thoughtful exchange with fellow panelists and for the engaged questions from students and industry participants. Encouraging to see conversations moving beyond tools and toward how organizations design better systems of decision-making. Giridhar Ramachandran Hari priya

80

2 Comments

Recommendations to the Industry by India AI Guidelines The proof of the pudding is in the eating as they say. Looking at the outcome of the India AI Guidelines released by MeitY on November 5, there were six guidelines to the industry players which is the net benefit to the industry. It must be noted that all Six guidelines are already being followed by FDPPI through its DGPSI-AI framework. unfortunately the literature research of the committee was not able to capture this information and add it in the references. Any way, all the recommendations are already under implementation under DGPSI-AI and there is nothing new. We are happy with the recommendations since they completely validate the DGPSI-AI framework though we do have to record our disappointment that no credit was given to efforts of the industry which was already ahead of the committee. Were they not aware? or Were they being silenced by vested interests?... will be known in due course. FDPPI Ministry of Electronics and Information Technology Balaraman Ravindran

15

3 Comments

Excited to welcome Sindhuja Gopalan and Sujit Kumar, PhD to our AI Science team at Solventum, led by Aravind Ganapathiraju. Sindhuja has over 13 years of experience in NLP and ML. During her 8-year tenure at Uniphore, she made impactful contributions across diverse NLP initiatives - ranging from traditional models to cutting-edge Generative AI applications - building scalable, production-grade AI systems along the way. Sujit comes from Nanyang Technological University (NTU), where he served as a Research Scientist specializing in Deep Learning and NLP. His work centers on LLMs, RAG and Multi-Agentic AI. His research contributions include misinformation detection, fact-checking, and developing LLM-powered reasoning, factuality, and AI-driven health coaching systems. A warm welcome to both as they join us on our journey of advancing AI innovation at Solventum!

81

6 Comments

🎮 Console Wars, Preferences & Science A Big Bang Take on Conjoint Analysis Last weekend, I watched The Big Bang Theory rerun where the gang debated which gaming console to buy. Chaos broke out. Sheldon ran simulations. Leonard weighed social factors. Raj asked for online multiplayer. And Howard? He just wanted the cheapest option. That’s when it hit me: this is the perfect setup to explain Conjoint Analysis! 🚀 What is Conjoint Analysis (in simple terms)? Conjoint analysis is a marketing tool used to figure out what people care about most when making choices between products. But instead of asking directly, “What’s more important to you: price or quality?”, we give people different combinations of product features — like: “Cheap but low quality” “Expensive but premium quality” And then ask them to rank, rate, or choose between them. This helps marketers uncover two key insights: ✅ Attribute importance — which features matter most? ✅ Part-worth utilities — how much each option level is liked (e.g., “free shipping” vs. “$10 shipping”) 🧠 The Big Bang Theory Analogy Imagine Sheldon, Leonard, Raj, and Howard are choosing a gaming console. Each console is defined by four key attributes: 🎮 Price (Cheap, Medium, Expensive) 🎮 Graphics Quality (Standard, HD, 4K) 🎮 Multiplayer Support (No, Yes) 🎮 Game Library Size (Small, Medium, Large) They’re shown several consoles with different combinations of these features: ConsolePriceGraphicsMultiplayerGame LibraryACheapHDYesMediumBExpensive4KYesLargeCMediumStandardNoSmall Each friend is asked to choose or rank their favorites. From their choices, we learn: 🧠 Sheldon values high-resolution graphics — he always picks the 4K option. 🧠 Leonard leans toward larger game libraries. 🧠 Raj only picks consoles with multiplayer support. 🧠 Howard? He always goes for the cheapest, regardless of features. So, instead of asking them what they want, we observe their trade-offs. That’s the beauty of conjoint analysis — it reveals what features people truly value when forced to choose. 🧬 Why “Conjoint”? Because we're analyzing attributes jointly — not one by one. It’s not “Do you like multiplayer?” but “How does multiplayer interact with price, graphics, and game library in your decision-making?” Summary (Big Bang Style): 🧪 If Sheldon were a marketing analyst: “Using conjoint analysis, I’ve deduced that a console with 4K graphics and multiplayer support maximizes consumer utility — assuming we can suppress Howard’s irrational love of discount pricing.” TL;DR: Next time you're designing a product, don’t just guess what people want. Use conjoint analysis — because people don’t just make choices, they make trade-offs. #MarketingScience #ConjointAnalysis #BigBangTheory #ProductStrategy #ConsumerBehavior #DataDrivenDecisions

2