Suzy Tonini’s Post

Salesforce and Galgotias University Launch Centre of Excellence in Tableau AI to Power Industry Ready Data Talent Greater Noida, 3rd April 2026: Galgotias University, in collaboration with Salesforce, the world’s leading AI CRM platform, has established the Salesforce Centre of Excellence – Tableau AI Data Lab on its campus. The initiative marks a significant step towards embedding Artificial Intelligence driven data analytics and visualization capabilities into mainstream academic learning, preparing students for the evolving demands of the data driven economy....

Salesforce and Galgotias University Launch Centre of Excellence – Tableau AI Data Lab to Build Industry Ready Data Talent Galgotias University, in collaboration with Salesforce, the world’s leading AI CRM platform, has established the Salesforce Centre of Excellence – Tableau AI Data Lab on its campus. The initiative marks a significant step towards embedding Artificial Intelligence driven data analytics and visualization capabilities into mainstream academic learning, preparing students for the evolving demands of the data driven economy.    Salesforce and Galgotias University launch Tableau AI Data Lab…...

AI upskilling isn’t optional — it’s your career advantage. Business leaders who combine AI literacy with practical application will shape the transformation ahead. Purdue Global School of Business and Technology programs are designed to build those capabilities: 🔹 BSBA Analytics — AI-powered data visualization and predictive insights 🔹 MSIT Core — machine learning, cybersecurity AI, and computational intelligence 🔹 Credential Stacks — CompTIA certifications paired with business strategy for hybrid leadership Start with literacy. Apply with confidence. Lead with impact. Ready to future-proof your career? Explore our programs: https://lnkd.in/e__4kDfq #AIUpskilling #BusinessIT #PurdueGlobal #PurdueGlobalSBIT #DataAnalytics

The data science training session is currently in progress at Tech365. Participants are not just learning tools. They are learning how to use data to solve real business problems and build intelligent systems. In the recent sessions, participants explored key foundations of data science, including statistics, probability, and machine learning concepts. These are the principles that power modern artificial intelligence systems and predictive analytics. Some of the areas covered so far include: • Understanding the role of statistics in machine learning • Descriptive and inferential statistics for data-driven decision making • Data distributions and measures such as mean, median, and variance • Detecting outliers and cleaning datasets using statistical methods • Probability concepts used in predictive models • Introduction to hypothesis testing and confidence intervals • Applying Python libraries to perform statistical analysis and model evaluation Participants also learned how these concepts are applied in real-world scenarios such as loan approval prediction, risk analysis, and automated decision systems. At Tech365, our focus is not just theory. We train participants to understand how data science actually works inside real organizations. By the end of the program, participants will be able to: • Analyze complex datasets • Build predictive machine learning models • Extract insights that help organizations make better decisions • Communicate data-driven results clearly This is what makes our training different. We focus on impact, practical application, and real-world problem-solving. If you want to build strong skills in Data Analytics, Data Science, Artificial Intelligence, Cloud, DevOps, Cybersecurity, or Software Engineering, Tech365 provides structured and hands-on training designed to make you globally relevant. #DataScience #MachineLearning #ArtificialIntelligence #DataAnalytics #TechTraining #Tech365 #FutureOfWork #AI #DataSkills

Building every machine learning model from scratch in 2026 is not thoroughness. It is an inefficient use of the most constrained resource in your data organization expert time. Enterprises are allocating months and significant compute budgets to training models on problems that the broader scientific community has already solved at scale. The foundation exists. The organizational awareness of how to leverage it often does not. Transfer learning is not a technical shortcut. It is a strategic acceleration mechanism that redefines what is achievable within a given budget cycle. What resource-intelligent data organizations have already embedded into their development philosophy: 📌 A pre-trained model fine-tuned on domain-specific data consistently outperforms a model trained in isolation on limited enterprise data 📌 The value of transfer learning is not in the model you inherit it is in the development cycles and capital expenditure you redirect 📌 Domain adaptation is the discipline that determines whether a transferred model performs or merely deploys The Chief Data Officer who requires every model to be built from first principles is not protecting quality. They are protecting a methodology that the industry has already moved beyond. Your data science teams are building from the ground up or building from the frontier down? #DataScience #MachineLearning #CDO

Iam glad to speak out of this award given to me by Nexford University under the completion of course of Data Sciences for Decision Making. Description: Data Sciences for Decision Making provides a foundation for learners to apply advanced analytics skills to complex data analysis and modes. Learners build an understanding of design, data analytics tools, and advanced analytics translators to communicate complex data-related requirements between data engineers, business, and IT stakeholders. Learners examine four major areas. First, there are momentum gains in the data and analytics revolution. Advances in data collection, machine learning, and computational power have fueled progress due to an exponential growth in the volume of data, continual advances in algorithms, and greater computational power and storage. Second, there are five elements of successful data analytics transformation: cases/sources of value, data ecosystems, modelling insights, workflow integration, and adoption. Third, the mapping value in data ecosystems includes data generation and collection, data aggregation, and data analysis. Fourth, models of distribution are fueled by big data analytics as business models are enabled by orthogonal data, hyper-scale, real time matching, radical personalization, massive data integration capabilities, data-driven discovery, and enhanced decision making.Data Sciences for Decision Making provides a foundation for learners to apply advanced analytics skills to complex data analysis and modes. Learners build an understanding of design, data analytics tools, and advanced analytics translators to communicate complex data-related requirements between data engineers, business, and IT stakeholders. Learners examine four major areas. First, there are momentum gains in the data and analytics revolution. Advances in data collection, machine learning, and computational power have fueled progress due to an exponential growth in the volume of data, continual advances in algorithms, and greater computational power and storage. Second, there are five elements of successful data analytics transformation: cases/sources of value, data ecosystems, modelling insights, workflow integration, and adoption. Third, the mapping value in data ecosystems includes data generation and collection, data aggregation, and data analysis. Fourth, models of distribution are fueled by big data analytics as business models are enabled by orthogonal data, hyper scale, real-time matching, radical personalization, massive data integration capabilities, data driven discovery, and enhanced decision making.

A new paper from Princeton challenges a core assumption in VLM training. Most people assumed the performance gap between open and proprietary models came down to better algorithms or more compute. But it's mostly about data diversity. --- The paper is Vero , published this week by researchers at Princeton University. The core idea: build a fully open Reinforcement Learning recipe for training vision-language models that works broadly — not just on one task type. --- What they built: Vero-600K — a 600,000 sample training dataset drawn from 59 open datasets, organized across 6 visual reasoning categories: → Chart & OCR → STEM reasoning → Spatial & Action understanding → Knowledge & Recognition → Grounding, Counting & Search → Captioning & Instruction Following Paired with task-specific reward functions (numeric, MCQ, bounding box IoU, click proximity, ordering, open-ended) and trained with a single-stage GSPO-based RL loop. No proprietary data. No staged training. Everything released. --- Results across 4 base models (all ~7-8B parameters): • Consistent +3.6 to +5.3 point improvement across 30 benchmarks • Vero-Qwen3I-8B outperforms Qwen3-VL-8B-Thinking on 23 of 30 benchmarks — without proprietary RL data • Vero-MiMo-7B outperforms MiMo-VL-7B-RL, which was trained on closed data from the same base model --- The finding worth paying attention to: Single-task RL does not generalize. Training only on STEM improves STEM but degrades grounding performance. Training only on grounding hurts captioning. The model develops qualitatively different reasoning styles per task — reflective and backtracking for STEM, direct perceptual search for grounding, structured synthesis for charts. These styles don’t transfer across categories. Broad coverage across task types, not algorithmic sophistication, is what drives strong RL scaling. --- Why this matters practically: For teams fine-tuning or building on top of VLMs, this suggests that curating diverse task coverage in your training data may matter more than optimizing your RL objective. And at the 8B scale, the performance gap between open and proprietary training pipelines appears to be closeable — with the right data. paper : https://lnkd.in/ghPT7s8D Github : https://lnkd.in/gun55fM3 #VLM #MultimodalAI #GenerativeAI #OpenSource #AIResearch #reinforcementlearning

The demand for Data Science and AI Engineering is skyrocketing, but the right certification can often be expensive. That’s where this opportunity comes in. 🚀 Through the IBM AI Engineering Professional Certificate, you can master the core of modern tech: ✅ Advanced Data Science ✅ Deep Learning & ML ✅ Portfolio Building The Best Part is the Government of Punjab, through PITB, will reimburse your certification fee once you successfully complete the program. It’s a full investment in your future at no cost to you. How to Apply: 1️⃣ Visit the portal: certifications.pitb.gov.pk 2️⃣ Select the IBM AI Engineering track. 3️⃣ Complete your certification and claim your reimbursement. Don't just learn about the data-driven future—lead it. 📊

Meet Stephen W., Data & AI Engineer and graduate from Andela’s intensive AI Engineering Bootcamp. Stephen specializes in SQL, Python, ETL Development, Data Architecture, Cloud Engineering, and Security. This AI Academy program fundamentally reshaped Stephen’s approach to engineering. He shared that he learned how to not only strategically embed generative AI into enterprise data workflows but also how this approach enables organizations to extract deeper value from their data assets while maintaining control, security, and compliance. For Stephen, AI adoption in organizations is not an option, it is fundamental. “AI adoption is no longer a differentiator but a necessity for sustainable growth. Enterprises that act now can reshape their data strategies around intelligent automation, predictive analytics, and contextual learning, creating a future-ready infrastructure that accelerates innovation and operational excellence. Waiting risks not only technological debt but also strategic irrelevance in increasingly data-driven markets.” If you’re building smarter systems and want to close the AI skills gap, find the AI ready talent like Stephen, here: https://lnkd.in/eT2ES_pq

Your data science team is running experiments daily. The institutional knowledge produced by those experiments is disappearing at the same rate. Machine learning development is an iterative discipline models are trained, parameters are adjusted, architectures are revised, and results are compared across dozens of configurations before a production candidate emerges. In most enterprise data science environments, that entire body of experimental work lives in individual notebooks, local file systems, and the memory of the engineer who ran the experiment last Tuesday. When that engineer moves teams, the organizational learning moves with them. Experiment tracking is not a developer productivity tool. It is the systematic preservation of the most expensive knowledge your data science organization produces. What reproducibility-mature data organizations have embedded as foundational infrastructure: 📌 An experiment that cannot be reproduced is not a result — it is an anecdote with computational provenance 📌 Hyperparameter logging without corresponding data versioning records half the experiment and creates the illusion of the whole 📌 The model that reaches production must be traceable to the exact conditions that produced it — anything less is a governance gap dressed as a deployment The Chief Data Officer who cannot answer which experiment produced the model currently governing their highest-stakes business process does not have a machine learning capability. They have a machine learning history that has already been lost. Your organization builds models does it retain the institutional knowledge of how those models were built? #DataScience #MLOps #EnterpriseAI