Science Forecasting Models

How do we bring AI to scientific modeling? The standard approach has been AI to augment existing numerical simulations. In a new work https://lnkd.in/gFMUvUbB we show this approach is fundamentally limited. In contrast, using the end-to-end AI approach of Neural Operators to completely replace numerical solvers helps overcome this limitation both in theory and in practice. Current augmentation approaches use AI as a closure model while keeping a coarse-grid numerical solver in the loop. We show that such approaches are generally unable to reach full fidelity, even if we make the closure models stochastic, providing them with history information and even unlimited ground-truth training data from full-fidelity solvers. This is because the closure model is forced to be at the same coarse resolution as the (cheap and approximate) numerical solver, and their combination does not result in high-fidelity solutions. In contrast, Neural Operators do not suffer from this limitation since they operate at any resolution and learn the mapping between functions. Neural Operators are first trained on coarse-grid approximate solvers, since we can generate lots of training data, and only use a small amount of expensive data from high-fidelity solvers in addition to physics-based losses to fine-tune the Neural Operator model for strong generalization. The key is that the Neural Operator model operates on any resolution, and can thus, accept data at multiple resolutions for training efficiently, without burdensome data-generation requirements.  Thus, Neural Operators fundamentally change how we apply AI to scientific domains.

𝐈 𝐣𝐮𝐬𝐭 𝐭𝐞𝐬𝐭𝐞𝐝 𝐬𝐨𝐦𝐞𝐭𝐡𝐢𝐧𝐠 𝐭𝐡𝐚𝐭 𝐬𝐡𝐨𝐮𝐥𝐝 𝐛𝐞 𝐨𝐧 𝐞𝐯𝐞𝐫𝐲 𝐟𝐮𝐭𝐮𝐫𝐢𝐬𝐭’𝐬 𝐫𝐚𝐝𝐚𝐫. MiroFish is an open-source AI prediction engine built by a Chinese undergraduate student in 10 days. It just topped GitHub’s global trending list, attracted 22,000+ stars, and secured a 30 million RMB investment from one of China’s most prominent tech founders. Here is why it matters beyond the hype. Most AI forecasting tools are still linear. You feed in data, you get a probability distribution, you call it a prediction. MiroFish does something structurally different. It builds a parallel digital world. You upload a “seed”: a news report, a policy draft, a financial signal. The system extracts entities, relationships, and social dynamics, then populates a simulated environment with thousands of AI agents, each with its own personality, memory, and behavioral logic. These agents interact freely. Emergent patterns surface. You observe the system evolve from a God’s-eye view, inject new variables, and watch trajectories shift. It is, in effect, a futures sandbox. A computational rehearsal space. For those of us who work in strategic foresight, this is not science fiction. This is the Futures Wheel and Causal Layered Analysis translated into a live simulation engine. The question we have always asked, “what happens next, and then what?”, can now be explored at a scale no human facilitation team could match. The use cases are immediately visible: policy stress-testing before legislation is passed, public opinion modeling before a brand launches, geopolitical scenario analysis, financial signal propagation in complex markets. Is it perfect? No. The quality of the simulation depends entirely on the quality and diversity of the seed data. Emergent behavior in agent-based models can also amplify biases present in the underlying LLMs. These are not small caveats. But the architecture is right. And the speed of adoption tells us the market agrees. If you are building in strategic intelligence, decision support, or AI-assisted foresight, this is a project worth studying closely. The future does not arrive as a single event. It emerges from millions of small interactions. MiroFish is the first open-source tool I have seen that takes that principle seriously at an engineering level. Link to the demo in the comments. What domains do you think benefit most from this kind of simulation-first approach to outcomes?

For those of you following developments with AI & Science, particularly around weather forecasting… At Google Research and Google DeepMind we have introduced an experimental model for tropical cyclone prediction, which can predict a cyclone’s formation, track, intensity, size and shape – generating 50 possible scenarios, up to 15 days in advance. And as we head into this year’s cyclone season, we’re partnering with the US National Hurricane center to support their forecasts and warnings. We’re publicly sharing this experimental model in Weather Lab, a new platform to access experimental weather forecast visualizations, and we hope to gather feedback and enable researchers and forecasters to leverage our models and predictions to inform their own work. You can learn more in our blog post (https://lnkd.in/geG62c2v) or this New York Times story (https://lnkd.in/gAFPbUrD).

Hot off our recent transformer paper, we're excited to share another AI model for precision medicine! Biological data collected from patients has exploded in recent years, presenting a challenge: how do we decipher that data to understand which patients will benefit most from specific therapies?  We in the Applied Data Science team at AstraZeneca are thrilled to share our paper in Cancer Cell called "AI-Driven Predictive Biomarker Discovery with Contrastive Learning to Improve Clinical Trial Outcomes." Here, we introduce the *Predictive Biomarker Modeling Framework (PBMF)*, a neural network-powered contrastive learning process that: 🔍 Explores vast multimodal datasets to uncover predictive biomarkers in an automated, systematic, and unbiased manner  🧠 Distinguishes predictive biomarkers (which indicate a likely benefit from a specific therapy) from prognostic biomarkers (which indicate general disease outlook)  💡 Distills its outputs into an interpretable decision tree, showing what drives treatment response In our studies, the PBMF:  📊 Surpassed existing methods in finding predictive biomarkers for immunotherapy success across various cancers in clinical trial and real-world data  📈 Discovered a predictive biomarker in an early-stage trial that boosted efficacy by 15% when retrospectively applied to the corresponding phase 3 clinical trial  📈 Discovered predictive biomarkers in single-arm early phase trial data with synthetic control arms, retrospectively improving the efficacy of the corresponding phase 3 trials by at least 10% We believe the PBMF has the potential to improve the way we design clinical trials and match patients to the right therapies. It can integrate with other models like our Clinical Transformer, creating exciting possibilities to someday discover biomarkers of adverse events, dosing strategies, and even to back-translate new drug targets. Read the full paper here: https://lnkd.in/eveAnVRY   Thanks to all the co-authors: Gustavo Arango, Damian Bikiel, Gerald Sun, Elly Kipkogei, Kaitlin Smith, Sebastian Carrasco Pro, Elizabeth Choe #PrecisionMedicine #ClinicalTrials #AIinHealthcare #Biomarkers #Immunotherapy

Inflation isn’t just an economic challenge—it’s a test of agility for businesses. As costs rise and purchasing power shifts, companies that rely on gut instinct risk falling behind. The real winners? Those who use data-driven insights to navigate uncertainty. 1️⃣ Understanding Consumer Behavior: What’s Changing? Inflation reshapes spending habits. Some consumers trade down to budget-friendly options, while others delay non-essential purchases. Businesses must analyze: 🔹 Spending patterns: Are customers shifting to smaller pack sizes or private labels? 🔹 Channel preferences: Is there a surge in online shopping due to better deals? 🔹 Regional variations: Inflation doesn’t hit all demographics equally—hyperlocal data matters. 📊 Example: A retail chain used real-time sales data to spot a shift toward economy brands, allowing it to adjust promotions and retain price-sensitive customers. 2️⃣ Pricing Trends: Data-Backed Decision-Making Raising prices isn’t the only response to inflation. Smart pricing strategies, backed by AI and analytics, can help businesses optimize margins without losing customers. 🔹 Dynamic pricing models: Adjust prices based on demand, competitor moves, and seasonality. 🔹 Price elasticity analysis: Determine how much a price hike impacts sales before making a move. 🔹 Personalized discounts: Use customer data to offer targeted promotions that drive loyalty. 📈 Example: An e-commerce platform analyzed customer behavior and found that small, frequent discounts led to better retention than infrequent deep discounts. 3️⃣ Demand Forecasting & Inventory Optimization Stocking the right products at the right time is critical in an inflationary market. Predictive analytics can help businesses: 🔹 Anticipate demand surges—especially in essential goods. 🔹 Optimize supply chains to reduce excess inventory and prevent stockouts. 🔹 Reduce waste in perishable categories like F&B, where price-sensitive demand fluctuates. 📦 Example: A leading FMCG brand leveraged AI-driven demand forecasting to prevent overstocking of premium products while ensuring budget-friendly variants were always available. 💡 The Takeaway Inflation isn’t just about rising costs—it’s about shifting consumer priorities. Companies that embrace data-driven decision-making can optimize pricing, fine-tune inventory, and strengthen customer loyalty. 𝑯𝒐𝒘 𝒊𝒔 𝒚𝒐𝒖𝒓 𝒃𝒖𝒔𝒊𝒏𝒆𝒔𝒔 𝒂𝒅𝒂𝒑𝒕𝒊𝒏𝒈 𝒕𝒐 𝒊𝒏𝒇𝒍𝒂𝒕𝒊𝒐𝒏𝒂𝒓𝒚 𝒑𝒓𝒆𝒔𝒔𝒖𝒓𝒆𝒔? 𝑨𝒓𝒆 𝒚𝒐𝒖 𝒖𝒔𝒊𝒏𝒈 𝒅𝒂𝒕𝒂 𝒕𝒐 𝒓𝒆𝒇𝒊𝒏𝒆 𝒚𝒐𝒖𝒓 𝒔𝒕𝒓𝒂𝒕𝒆𝒈𝒚? 𝑳𝒆𝒕’𝒔 𝒅𝒊𝒔𝒄𝒖𝒔𝒔 𝒊𝒏 𝒕𝒉𝒆 𝒄𝒐𝒎𝒎𝒆𝒏𝒕𝒔! #datadrivendecisionmaking #dataanalytics #inflation #inventoryoptimization #demandforecasting #pricingtrends

To create good policy you need responsible foresight, enabling ethical, sustainble, accountable future design. AI now can massively enable human-centered responsible foresight, in helping address uncertainty, assess risks, and set policies for creating better futures. María Pérez Ortiz's new paper "From Prediction to Foresight: The Role of AI in Designing Responsible Futures" describes responsible foresight in policy and the role of computational foresight tools. Notable approaches to using AI in responsible foresight include: 🤝 Participatory Futures for Inclusive Planning. Engaging diverse stakeholders in foresight practices democratizes the future-planning process. AI tools streamline public participation by analyzing preferences, simulating collective decisions, and creating urban plans that reflect community values, fostering equity and resilience. 🧠 Superforecasting for Precision and Insight. Superforecasting uses disciplined reasoning and probabilistic thinking to predict uncertain events. AI-powered assistants improve human forecasting accuracy by 23%, aggregating data and refining predictions through collective intelligence and advanced analytical models. 🌐 World Simulation for Systemic Insights. Advanced modeling frameworks simulate interconnected global systems, enabling policymakers to test "what-if" scenarios. AI accelerates these simulations, providing precise forecasts and dynamic platforms to visualize the long-term consequences of policy decisions across economic, social, and environmental domains. ⚙️ Simulation Intelligence for Decision Optimization. By integrating AI with high-fidelity simulations, simulation intelligence explores complex systems to uncover optimal strategies. This tool assists in crafting effective policies for urban planning, sustainable agriculture, and climate resilience, offering actionable pathways for addressing systemic challenges. 📜 AI-Assisted Narrative Techniques. Large language models contribute to speculative futures by generating detailed "value scenarios" that integrate ethical, technological, and societal considerations. These AI-driven narratives enable policymakers to visualize desirable outcomes and evaluate potential trade-offs. 🔗 Hybrid Intelligence for Enhanced Foresight. Combining human creativity with AI’s computational strengths creates a robust foresight framework. Intuitive interfaces, explainable AI, and participatory design ensure that tools remain transparent and aligned with ethical considerations, empowering policymakers to navigate complex challenges collaboratively. ♻️ Iterative Foresight with Feedback Loops. Continuous monitoring and real-time adaptation enhance foresight processes. AI’s ability to process evolving data and generate actionable insights ensures policies remain responsive, flexible, and aligned with long-term objectives. The power of AI in assisting foresight is just beginning to come to fruition.

New paper – A foundation model for the Earth system Abstract “Reliable forecasting of the Earth system is essential for mitigating natural disasters and supporting human progress. Traditional numerical models, although powerful, are extremely computationally expensive. Recent advances in artificial intelligence (#AI) have shown promise in improving both predictive performance and efficiency, yet their potential remains underexplored in many Earth system domains. Here we introduce Aurora, a large-scale foundation model trained on more than one million hours of diverse geophysical data. Aurora outperforms operational forecasts in predicting air quality, ocean waves, tropical cyclone tracks and high-resolution #weather, all at orders of magnitude lower computational cost. With the ability to be fine-tuned for diverse applications at modest expense, Aurora represents a notable step towards democratizing accurate and efficient Earth system predictions. These results highlight the transformative potential of AI in environmental forecasting and pave the way for broader accessibility to high-quality #climate and #weather information.” Bodnar, C., Bruinsma, W.P., Lucic, A. et al. A foundation model for the Earth system. Nature 641, 1180–1187 (2025). https://lnkd.in/eh8wQ2wx

How Zinnia Used AI to Forecast Daily Call Volumes with 95% Accuracy 📞 At Zinnia, we needed a better way to forecast call center volumes — our existing tool often missed the mark by 10–20%, making staffing plans unreliable So, we rolled up our sleeves and built our own AI forecasting solution: ✅ Combined Prophet (seasonality + trends) with XGBoost (learn from errors) ✅ Used real-world signals like holidays, month-ends, and even Mondays after weekends ✅ Tuned everything with time-aware cross-validation We tested A LOT (even LSTMs and SHAP-based pruning!), but our hybrid model consistently delivered 95%+ accuracy across clients. 🔍 I’ve shared the full breakdown, code, and what worked (and what didn’t) in this Medium article — practical, real-world AI for ops. If you're a data scientist, ML engineer, or even an ops leader — this one’s for you. Josh Everett | Pawan Choudhary | Daniel Gremmell | Eti Gupta #DataScience #Forecasting #AI #XGBoost #Prophet #TimeSeries #MLinProduction #CallCenterAI #WorkforcePlanning #ZinniaTech #AIinOps

Physically Consistent Forecasting Critical time series cannot rely on statistical correlation alone. When regimes shift, purely data-driven models collapse. In real systems, physics remains valid. Technical Architecture - Input Multivariate historical data + structural physical variables - Pinneaple Pipeline: . pinneaple_timeseries (temporal engineering and physics-aware feature preparation) . pinneaple_models (hybrid backbone: temporal dynamics + physical structure) . pinneaple_solvers (structural constraints and conservation laws enforcement) . pinneaple_train (combined loss: predictive error + physical violation penalty) Output Robust forecasts with stability under distribution shifts. Use Case Load forecasting in electrical grids under extreme events. Heat waves, substation failures, abrupt consumption spikes. Purely statistical models extrapolate poorly. The hybrid model: - Learns temporal patterns - Respects system physical limits - Penalizes structural violations during training - Maintains stability under extreme regimes The Principle - pinneaple_models defines the hybrid dynamics. - pinneaple_train enforces structural consistency during learning. When statistics fail, physics anchors the model. This is not just forecasting. It is constraint-aware engineering.

What if your hospital could predict a crisis… before it happens? Here’s how one mid-sized hospital turned used our predictive analytics model in their system. 📍Background: A 200 bed multi specialty hospital in Tier 2 India was constantly under pressure. Stockouts of critical medicines Sudden patient surges with no staff planning Equipment lying idle in one department while another faced shortages Finance team always firefighting Revenue was falling. Patient care was inconsistent. Staff was burning out. They implemented a Predictive Analytics System linked to: Patient admission history OPD trends Seasonal disease patterns Staff rosters Inventory data Billing + discharge cycles Within 3 months, the dashboard could show: 1) Which departments will have a spike next week 2) Which medicine stocks will run out in 10 days 3) How long each patient stays, on average, for each treatment 4) Where staffing gaps will occur in coming shifts 5) Where revenue leakages were happening due to idle assets The Impact: - Improvement in inventory efficiency - 31% drop in emergency stock orders - Higher staff availability during peak hours - Reduced patient wait time by 26% - Cost savings of ₹1.8 crore/year Predictive Analytics helps hospital leaders move from reactive mode to proactive control. It’s how hospitals stop surviving and start scaling. Whether you're managing a single unit or a hospital chain, Start by asking: "What patterns am I missing in my daily operations?" Because in healthcare, even a 1% smarter decision can save a life. Agree? #HealthcareInnovation #Predictiveanalytics #Hospital #tech