Ankur Undhad

Ever feel like today’s AI models are getting smarter but somehow less useful? You’re not imagining it. LLM benchmarks have become the obsession. Model makers are locked in a cycle of chasing leaderboard points on MMLU, GSM8K, and other tests no one actually cares about outside of research labs. The result? Models that ace exams but fall apart in real-world use. They misinterpret vague instructions. They forget what you said three messages ago. They can’t think beyond a single step. But sure, the score went up by 0.3 percent. Benchmarks were supposed to be a guide. Instead, they’ve become a crutch. A way to show “progress” without improving anything that matters. The questions we should be asking are simple: Can the model follow messy, human instructions? Can it reason through problems, not just regurgitate answers? Can it actually help people get real work done? If you’ve worked with these models, you’ve felt the gap. You’ve seen the hype, and then you’ve seen it fall short when it counts. Are we building for people, or are we building for points? I’d love to hear what others are seeing in the wild. #AI #LLMs #Benchmarks #MachineLearning #RealWorldAI

27

3 Comments

AI Revolutionizes ITeS Industry! The ITeS and BPM sectors aren't dying; they're being reborn through AI. We're moving from FTE models to outcome-based AI, where agentic AI is the new workforce. This isn't just automation; it's strategic innovation driving the next wave of growth. #ITeS #AIRevolution

2

After 10 years in data science, it’s rare to come across a shift this significant—AI agents are no longer just an idea, they’re being deployed. Unlike traditional models that predict, these agents act. They interpret context, connect across systems, and execute tasks—autonomously. From underwriting and reporting to customer support and operations, businesses are using them to drive real outcomes. This isn’t about replacing humans—it’s about freeing us from the repetitive, so we can focus on strategy, insight, and impact. As someone who’s spent a decade building data solutions, I see this as the next chapter: moving from dashboards to decision-making, from insight to intelligent action. We’re not just building models anymore. We’re building systems that think. #AI #DataScience #AIagents #Automation #Innovation #FutureOfWork

18

6 Comments

The North Texas AI community continues to lead practical, implementation-focused discussions bridging theory with real-world results. Recent IoT meetup in North Texas brought together technology leaders, manufacturing experts, and AI practitioners for deep-dive discussions on enterprise AI Participants (Partial List): Walker Reynolds - IIoT & Industry 4.0 Solutions Architect Tom Hamilton - Telecom Infrastructure & Inventory Optimization Alex Arias (Texas Instruments) - Smart Market Systems Engineer Nicholas Yarmesch (Uplink) - IoT Deployment Platform Founder Hubert Rhasaj (Oyap) - Service Business Digital Transformation Mohammed Hassan - Energy Systems & Battery Management Developer Carl Daley (Valueyetics) - AI Consultancy Founder Sergey Rossi - Predictive Maintenance Solutions (25+ years experience) Ron Assumption - AI Copilots for Wireless Networks Co-Founder Gavin Lehotsky (Texas Automation Systems) - Plant Automation Specialists Key Demonstrations & Case Studies: Walker showed a live demonstration of: UNS Studio - 6.5M line IIoT platform enabling real-time manufacturing dashboard generation in minutes rather than weeks Model, Context, Protocol (MCP) framework showcase - demonstrating AI agents interacting with existing enterprise systems Bible application development case study - 160K lines of code completed in 4 days with Cursor The discussions emphasized three critical themes: Data Revolution: Moving from unreliable manual data collection to AI-supervised precision Speed-to-Value: Dramatic reduction in development cycles through agentic AI implementation Human-AI Synergy: Leveraging AI for execution while preserving human advantages in creativity, empathy, and strategic thinking Key Takeaways: Data fidelity breakthrough: AI agents achieve 99.999% accuracy when human-supervised vs. 98% error rate in traditional collection - using agentic AI - achieving 99.999% data fidelity vs. traditional 2% human accuracy (98% "pencil-whipped") Development acceleration: Complex applications now deployable in days rather than months through agentic frameworks Industry resistance reality: Manufacturing adoption remains slow despite proven ROI, citing trust and cultural barriers Employment evolution: Focus shifting from replacement fears to skill augmentation and new role creation Energy implications: Nuclear power sentiment shifting positive among environmentalists as AI demands increase Human differentiation: Emotion, intrinsic motivation, and empathy remain uniquely human competitive advantages These session highlighted that we're entering an unprecedented era where AI could create true abundance for the first time in human history, while emphasizing that success depends on thoughtful human-AI collaboration rather than wholesale automation. #AI #Manufacturing #IoT #AgenticAI #DigitalTransformation #DataFidelity #TechLeadership #Innovation

18

2 Comments

Most #RAG performance issues are diagnosed as retrieval failures when they are actually preprocessing failures. A deep dive into #OCR mechanics reveals a hard ceiling on accuracy that no amount of #promptengineering can fix. The degradation comes in two specific forms. 1. #Semanticnoise. When character-level errors occur. For example, a zero is read as the letter 'O' or a #Banking term is misspelled, the #vectorembedding shifts entirely. The relevant #chunk is never retrieved because it mathematically no longer resembles the user's query. 2. More costly for enterprise data, is formatting noise. This effectively destroys tables. Tables rely on #spatial relationships, not #sequential text. When OCR flattens a financial table into a linear string, the link between a specific row header and a cell value is often severed. Tests show this causes a 50% #degradation in answer quality, regardless of whether the underlying model is #GPT5 or a smaller open-source alternative. The most efficient architecture to solve this is not a single super-model, but an adaptive routing strategy. #Pipeline approaches (like #Docling / #MinerU / #Marker) are utilized for preserving layout structure. End-to-end models are targeted at scientific formulas. #VLMs are reserved for complex charts and infographics where visual context is mandatory. #AIML #GAN #AE #AIEngineer #GenAI #DataScience

1

AI is shifting from single models to ecosystems of specialized agents. This research on Reinforced Agent Merging (RAM) highlights a critical reality:  If you merge RL-trained agents the same way we merged fine-tuned models, you destroy value. Why it matters:  • RL agents learn sparse, task-specific intelligence  • Naive averaging dilutes the very behaviors you trained them for  • “One model to rule them all” is no longer viable What RAM shows instead:  • Share common intelligence  • Preserve and amplify unique capabilities  • Combined systems can outperform individual specialists Impact for leaders and builders:  • AI platforms will be agent-based, not model-centric  • Competitive advantage shifts to orchestration and integration  • The hardest problem is no longer training models, but merging intelligence without loss This is where AI strategy meets product leadership. If you are still designing around a single assistant, you are already behind. #AI #ArtificialIntelligence #AgenticAI #ReinforcementLearning #GenAI #AIArchitecture #AIProductManagement #AILeadership #EnterpriseAI https://lnkd.in/ed7pTrsD

7

I've spent a good portion of my career in data science and predictive AI, but the rise of Generative AI and Agentic AI feels truly transformative. While early Generative AI focused on friendly uses, Agentic AI is now making complex enterprise applications genuinely feasible. Despite this progress, one truth remains constant: quality, well-governed internal data is crucial. Without it, even the most innovative models are limited. This shift is evolving the role of the AI leader. It’s no longer just about algorithms; it’s about being a champion of data, a strategist of systems, and an advocate for ethical, collaborative AI. This involves fostering trust, building data literacy, and aligning AI initiatives directly with business return on investment (ROI). To achieve impactful enterprise AI, two key practices are essential: First, building a sustainable AI technology operating model with clear roles, governance, and targets. For a deeper dive, I recommend Technology Operating Models for Cloud and Edge (https://lnkd.in/ghSkNM6Q) by Ahilan Ponnusamy and Andreas Spanner. Second, implementing an AI platform based on platform engineering principles and operated as a product. It needs to be continuously evolving, secure, and desirable to use. Ultimately, successful enterprise AI comes from alignment: aligning data, strategy, platform, people, and processes around a shared vision. AI leaders who prioritise this will not only deliver impact today but also help architect a future where AI is truly integrated and trusted. Please let me know your thoughts. Are we building intelligent and accountable enterprise AI, or just beginning this journey? #AI #GenerativeAI #AgenticAI #DataScience #Leadership #EnterpriseAI #DigitalTransformation #AILeadership #DataStrategy #PlatformEngineering #AIOperatingModel #AIEthics #DataCulture

22

5 Comments

The Reality of AI Model Trimming in Manufacturing Your AI model achieves 99% accuracy in the lab. Then you try to deploy it on the factory floor—and reality hits hard. — Why Trimming is Critical — Manufacturing environments don’t forgive bloated models: • Edge constraints where processing power is measured in watts, not teraflops • Real-time demands where a 2-second delay halts entire production lines • Cost pressures where compute costs multiply across hundreds of sites • Reliability needs where model failure means safety incidents — The Trimming Challenge — Model optimization isn’t just about size—it’s surgical decision-making: - Performance vs. speed - which accuracy can you sacrifice for faster inference? - Memory vs. processing - smaller models or faster computations? - Feature importance across different production scenarios - Robustness preservation - ensuring trimmed models handle edge cases The hardest question: “What can we afford to lose without losing what matters?” The Critical Partnership: Data Scientists + Field Engineers! This collaboration is essential: Data Scientists bring expertise in pruning, quantization, and architectural optimization. Field Engineers provide production reality: • Which delays are tolerable vs. catastrophic? • What accuracy thresholds maintain quality? • Where can we accept false positives vs. negatives? — Success Patterns — The most successful deployments happen when: • Engineers define real-world benchmarks upfront • Data scientists optimize iteratively based on production constraints • Teams test under real manufacturing conditions • There’s feedback between deployment performance and refinement The Bottom Line AI model trimming for manufacturing isn’t a technical exercise—it’s a business-critical collaboration. The models that succeed in production are those built by teams that understand both the mathematics of optimization and the realities of manufacturing operations. What’s been your experience with deploying AI models in resource-constrained environments? How do you balance model performance with practical deployment requirements? #ManufacturingAI #ModelOptimization #EdgeAI #IndustrialAutomation #AIDeployment #Manufacturing #ProductionAI #AIOptimization

19

2 Comments

🏀 Masterstroke of Creating Chaos - Do H1B hurts Indians? 🏀 OPINION DIFFERS Recent reports indicate that a new U.S. presidential proclamation has significantly altered the H-1B visa landscape, particularly impacting Indian professionals who are the largest group of beneficiaries. Here's a summary of the key "threats" and the current situation: 🌀 The New H-1B Visa Fee The most significant change is the introduction of a steep, one-time fee of $100,000 for new H-1B visa petitions. This is a massive increase from the previous fees, which typically ranged from $2,000 to $5,000. 🌀 Who it affects: This fee applies to new H-1B applications, including those submitted in the next lottery round. It does not apply to current H-1B visa holders or visa renewals. 🌀 Who pays: The fee is to be paid by the sponsoring company, not the individual applicant. 🌀 Purpose: The stated goal of this policy is to limit the program to "very highly skilled" professionals and to prevent companies from using the visa to bring in lower-paid, entry-level workers who could be replaced by American citizens. 🌀 Immediate and Potential Impacts on Indians Reduced opportunities for new applicants: The exorbitant fee makes it economically unfeasible for many companies, especially smaller firms and startups, to hire new H-1B workers. This could drastically reduce the number of new visas issued and limit the career opportunities for Indian professionals seeking to work in the U.S. 🌀 Disruption and uncertainty: The sudden announcement of the policy caused panic and confusion among Indian H-1B visa holders who were outside the U.S. at the time. Despite clarifications that the fee does not apply to existing visa holders, many rushed to return to the U.S. to avoid any potential issues, leading to chaos at airports and a surge in flight ticket prices. 🌀 Impact on Indian IT companies: Indian IT firms are major users of the H-1B visa program. The new fee is expected to significantly increase their operational costs and may force them to rethink their business models, potentially leading to a shift toward more offshore projects. 🌀 Concerns for families: The change can cause significant disruption for families. Spouses and children on H-4 visas are tied to the H-1B visa holder's status, and the uncertainty surrounding the new rules has created a great deal of anxiety. 🌀 India's Response and Potential "Silver Linings" Some experts in India see a potential silver lining in the new policy, suggesting it could lead to a "reverse brain drain." The argument is that if skilled Indian professionals cannot easily go to the U.S., they will return to India, which could boost the domestic economy and accelerate the growth of Global Capability Centers (GCCs) and the Indian technology sector as a whole. Stephen D J A #h1b #visa #visaimpact #Wednesdaytopic

4

1 Comment

The world is steadily shifting from task-based automation to intelligent orchestration — powered by GenAI. GenAI is no longer a proof of concept; it's becoming a core pillar in how we design workflows, decisions, and experiences. As language models evolve into agents, and search becomes reasoning via Retrieval-Augmented Generation (RAG), we’re seeing the emergence of truly AI-native operating models — especially in domains like finance, compliance, and operations. Staying close to this evolution means staying hands-on. Recent explorations — from experimenting with agentic workflows to evaluating output quality using RAG pipelines — have offered a glimpse into what “real-world-ready” GenAI looks like when applied in production-grade financial contexts. Some of these learnings translated into practical experiments — a few of which I’ve shared through posts and prototypes: 🔍 LIC Policy Document QA using RAG (https://lnkd.in/gsqZn2gV) 🤖 The Rise of Reasoning Machines (https://lnkd.in/gj79FmmV) 🧠 Explainability: From Prompting to Agents (https://lnkd.in/gXzwwa7V) 📊 GenAI for SEC Filing Insights (https://lnkd.in/gv85dBnJ) ⚙️ No-Code App Prototypes with GenAI (https://lnkd.in/gJ5VqJSC) Recent certifications have been very helpful in this ongoing journey. 🎓 LangChain for LLM Applications – Analytics Vidhya https://lnkd.in/g9ndgF8J 🎓 Generative AI – Great Learning A foundational program that helped set the stage for deeper exploration of LLMs, prompting strategies, and use-case development. Also currently reading Hands-On Large Language Models by Jay Alammar — an excellent guide that combines conceptual clarity with engineering intuition for building with LLMs. The learning never stops — especially when the landscape is evolving this fast. #GenAI #AIProductLeadership #AgenticAI #LangChain #AnalyticsVidhya #GreatLearning #LLM #RAG #EnterpriseAI #AIinFinance #AITransformation #LLMEngineering

30

1 Comment