DataSeer: AI-powered solutions for open research

📢 Exciting Collaboration Alert! VS Software Lab is thrilled to announce our MoU with VPKBIET, Baramati – Department of AI & DS! 🤝 This partnership will empower students with hands-on training, real-world AI & Data Science exposure, and industry insights, bridging the gap between academics and practical tech skills. Together, we’re shaping the next generation of tech innovators! 🚀 #VSSoftwareLab #VPKBIET #AI #DataScience #IndustryCollaboration #TechEducation #Innovation #FutureReady

OpenAI recently introduced GDPval, a new benchmark designed to measure how well AI models perform on economically valuable, real-world tasks. Unlike traditional academic benchmarks, GDPval focuses on realism and relevance, evaluating models across 44 knowledge work occupations (from software developers and lawyers to nurses and financial analysts) with tasks crafted by seasoned professionals and graded by industry experts. Some of my key takeaways were: 1) Real-world scope: Tasks reflect actual deliverables like legal briefs, engineering designs, and care plans, not just exam-style questions. 2) Performance progress: Frontier models are now approaching human-level quality, with Claude Opus 4.1 and GPT-5 leading the pack. 3) Efficiency gains: Models complete tasks 100x faster and cheaper than humans (though oversight remains critical). 4) Future of work: AI is poised to take on repetitive, well-structured tasks, freeing people to focus on creativity, judgment, and higher-value decision-making. GDPval is still in its early stages (limited to one-shot evaluations today), but it marks a significant move toward more realistic benchmarks that reflect the true complexity of professional work. Linking the official documentation in the comment below. #evaluation #llm #artificialintelligence #futureofwork #benchmarks

Super exciting news from Thermo Fisher Scientific‼️ — we’re partnering with OpenAI to help scientists and researchers bring life-changing medicines to patients faster and more efficiently. By embedding OpenAI’s advanced capabilities across our operations — from product development to clinical research — we’re enhancing innovation, productivity, and impact. https://lnkd.in/gZBNcDB3 #ThermoFisherScientific #OpenAI #AI #Innovation #LifeSciences #ServingScience

What if performing sophisticated open science editorial screening were faster, more consistent, less demanding, and more affordable? DataSeer SnapShot makes it possible. SnapShot is an AI technology solution developed specifically for and by scholarly journal editors. It works with editors, automating the tedious aspects of research integrity checks, freeing your team to focus on the work that requires their expertise and creativity. Explore DataSeer SnapShot ➡️ https://lnkd.in/eNEAZyV9

Anthropic has launched Claude Haiku 4.5, a new version of its scaled-down AI model. The model offers similar performance to Sonnet 4 at one-third the cost and more than twice the speed, according to the company. Anthropic's CPO Mike Krieger said the model opens up new deployment possibilities, allowing for complex planning with Sonnet and fast execution with Haiku. The model is expected to be particularly useful in software development tools, offering high intelligence and speed for real-time tasks. Read more: https://lnkd.in/e5Z-sKie 📰 Subscribe to the weekly AI Programming Weekly: https://lnkd.in/eUQ-KTc2 #ai #artificialintelligence #ainews

“Knowledge graphs are super hot right now.” Multiple people said this at our booth at BioTechX Europe this week, and it stuck with me. Not just because it’s catchy, but because it’s true. We at ONTOFORCE feel the growing interest in knowledge graphs every day. In an era of data overload, knowledge graphs offer a meaningful way to connect, contextualize, and reason over complex biomedical data. So why the sudden heat around something that’s been around for a while? Here’s what’s driving it: 🧠 Semantic layer is the game-changer At the core of a knowledge graph is a semantic layer, an ontology-driven structure that doesn’t just store data, but understands relationships. It allows machines (and humans!) to query information contextually, not just syntactically. 🔗 It’s not just data ; it's linked knowledge In life sciences, you’re often pulling from genomic data, clinical trials, literature, patient records, and more. A knowledge graph connects these dots in a machine-readable way. Suddenly, questions like “Which genes are implicated in both disease A and B?” become easily answerable. 🤖 AI loves context Large language models (LLMs) and other AI tools can generate better insights when grounded in structured, contextual data. Knowledge graphs provide exactly that: a layer of meaning that makes AI outputs more trustworthy and interpretable. 📈 From discovery to decision Whether you’re in drug repurposing, target identification, clinical trial design or regulatory, knowledge graphs accelerate the journey from scattered information to informed action. 💡 TL;DR: Knowledge graphs are hot because they make data intelligent. And in life sciences, that intelligence is the difference between information and innovation. #BioTechX #KnowledgeGraphs #SemanticAI #LifeSciences #DrugDiscovery #BiomedicalInformatics #GraphTechnology #ONTOFORCE

Unlock the future of secure software development with Federated Learning. Discover how decentralized data science and privacy-preserving machine learning are transforming AI in software development, across healthcare, finance, and beyond. Read the full blog to see how your business can innovate securely: https://lnkd.in/gFYXC_hr #InApp #FederatedLearning #DataScience #AI

Agentic AI may still be maturing when it comes to solving complex, large-scale business challenges, but it’s already showing strong value for smaller, everyday problems. My Agentic AI assistants support me in learning, personal finance, and even hobbies😊 Here’s a sample report generated by an Agentic AI workflow in LangGraph, where an LLM (gpt-5-nano) handled the full process: Planning and execution->Calling search tools->Summarizing data (gemini-2.5-flash) ->Creating the markdown report->Exporting it as a PDF. Exciting times ahead!! #ArtificialIntelligence #AgenticAI #GPT5 #GeminiAI #FutureOfWork

When is data "good enough"? During one of my interviews with a tech provider, we got into a rich discussion about data quality. His answer began with something I keep returning to: -All these technologies (computer vision, natural language processing, artificial intelligence) are statistical technologies. And to make correct statistics, you need perfect data. But perfection doesn’t exist. So you need "good enough data" to make estimations, not truths. He described how in agriculture, we often work with images, each pixel a variable, sometimes over a million pixels per image. "To do statistics with a million independent variables" he said, "the complexity of the problem becomes very, very big." But what really stayed with me was how he separated the scientific and the business perspective on what matters: - In business, high-quality data means stable conditions, repeatable patterns that make predictions reliable. You constrain variability so that your model can perform with 99% accuracy in a clearly defined scope. You aim for consistency, because that’s what enables you to deliver results, products or services that actually work in the real world. - But in science, you seek the opposite. You invite variability. You want to test your models under every possible condition (different seasons, devices, angles, lighting or environments) to build something more general, robust and transferable. You aim for diversity, because that’s what allows you to uncover deeper truths and general principles. It’s the same question seen through two lenses. Maybe "good data" in agriculture is simply where the two meet, consistent enough to work, variable enough to matter #PhDResearch #AgriculturalData #DataValue #EnTrustProject #QualitativeResearch #AgriData #DataSharing #DigitalFarming #EnTrust #MSCA #SmartAgriculture #PhDProcess #EnTrustdn #DataInAgriculture #DigitalAgriculture #New_Generation_of_Data_Executives #ResearchImpactEU #SingularLogicRnD #AUA

🔎 Why do LLMs struggle with long contexts—and what can we do about it? Just read a fascinating new paper: ChunkLLM: A Lightweight Pluggable Framework for Accelerating LLMs Inference. The authors tackle a huge bottleneck in LLMs: the quadratic computational cost of self-attention. ChunkLLM solves this with modular adapters and input chunking, achieving efficient inference with minimal resource usage and retains strong semantic performance. This plug-and-play framework means faster, scalable deployments for AI practitioners with clear experimental results—especially for long-context, high-token applications. Would you incorporate ChunkLLM in your projects? Full paper: https://lnkd.in/drhBhkcQ #AIResearch #LLM #MachineLearning #DeepLearning #AIEngineering