KI Bowle Event – This Wednesday! 🍹 How do you bridge Safety and AI? Join us this Wednesday for our next KI Bowle, where Dr. Tim Fischer from SICK will be sharing his insights on Safety and AI. SICK is a world-leading manufacturer of intelligent sensors and sensor solutions for factory, logistics, and process automation. A must-attend for anyone interested in robotics, industrial automation, or the future of intelligent systems. In-person event, Linz | A few seats still available! Register here: https://lnkd.in/ddJE78H7
NXAI
Technology, Information and Internet
Linz, Upper Austria 7,511 followers
Accelerating Industrial AI
About us
At NXAI, we are dedicated to unlocking new AI verticals, pioneering solutions that transform manufacturing, engineering, logistics, and energy sectors through scalable AI architectures and simulations.
- Website
-
https://www.nx-ai.com/
External link for NXAI
- Industry
- Technology, Information and Internet
- Company size
- 11-50 employees
- Headquarters
- Linz, Upper Austria
- Type
- Public Company
- Founded
- 2023
- Specialties
- Large Language Model, xLSTM, KI, Technologie, Forschung, Künstliche Intelligenz, Artificial Intelligence, Industry, AI, and Simulation
Locations
-
Primary
Get directions
Peter-Behrens Platz 2
Linz, Upper Austria 4020, AT
Employees at NXAI
Updates
-
Welcome (back) to NXAI, Leo! 🦖 After a break in Asia, we are officially welcoming Leo Sollereder back at NXAI and are excited to announce that he will be joining our Management Team as our Founders Associate. With a software development background, Leo has evolved into the VC and startup scene over the years. He also completed his studies at WU Vienna, specializing in Computer Science with a focus on Data Science and Entrepreneurship & Innovation. As Founders Associate at NXAI, Leo will work directly with our founders, solving critical problems across the company and helping the team move faster. We are really glad to have you on board, Leo! 🚀
-
-
xLSTM proves its worth in the medical sector ❤️
⏳ 30 Days to Go: The "Lost in Time" Lineup is officially live! We are exactly one month away from June 10th. Kora Helm and I are excited to reveal the preliminary program for our symposium in Augsburg. We are moving beyond "snapshots" toward a regional network for temporal clinical data. Our speakers will dive into three high-impact sessions: · Session 1: ML Methods – Including Scientific ML, confidence-aware prediction via xLSTM, and few-shot pathology classification. · Session 2: Electrical & Physiological Signals – From cardiac output estimation to mortality prediction and dynamics in consumer smart-watches. · Session 3: Optical & Imaging-Based Biosignals – Featuring camera-based vital signals and temporal dynamics in longitudinal imaging. Why join us? This is a founding meeting for a new regional network. We have dedicated time for a catered networking reception to spark new collaborations. 📍 Location: IHK Schwaben, Augsburg (Bartholomäus-Welser-Raum) 🗓️ Date: June 10, 2026 | 13:30 – 17:45 🎟️ Registration: Free of charge (Supported by B. Braun-Stiftung, Universit��t Augsburg and Universitätsklinikum Augsburg). Don’t get lost in time—secure your spot today. Registration link in the first comment! 👇 #MedicalAI #TimeSeries #Augsburg #DigitalHealth #ClinicalDataScience #Innovation #LostInTime
-
-
🤴Thanks for sharing
My essay today at The Sequence: Return of the King: Unrolling the xLSTM Architecture: https://lnkd.in/eqd6pqWr
-
Again xLSTM 🚀
Excited to share that two of our conference papers have been officially published in IEEE Xplore! 🎉📄 ━━━━━━━━━━━━━━━━━━━━━━━━━━━ 📄 Paper 1: "Multiclass Emotion Classification using xLSTM Learning of Optimized Features from EEG Signal" Published in the 2025 IEEE International Conference on Biomedical Engineering, Computer and Information Technology for Health (BECITHCON). 🔗 https://lnkd.in/gdA6cXry Key contributions: ✅ High-dimensional feature optimization: 2549 → 200 features via mRMR ✅ xLSTM architecture with mLSTM & sLSTM for long-term dependency modeling ✅ Exponential gating & memory mixing for enhanced sequential learning ✅ 99.06% accuracy — outperforming conventional LSTM (97.17%) ━━━━━━━━━━━━━━━━━━━━━━━━━━━ 📄 Paper 2: "Hand-xLSTM: A Deep Learning Framework to Classify Hand Movements Motor Imagery EEG" Published in the 2025 IEEE 4th International Conference on Robotics, Automation, Artificial-Intelligence and Internet-of-Things (RAAICON). 🔗 https://lnkd.in/gg3zsTg5 Key contributions: ✅ Alternating sLSTM & mLSTM architecture for sequential & nonlinear EEG dynamics ✅ Frequency band decomposition across 5 EEG channels → 25 features/sample ✅ SHAP-based explainability — beta & alpha bands as dominant contributors ✅ 98.72% accuracy, surpassing prior state-of-the-art approaches ━━━━━━━━━━━━━━━━━━━━━━━━━━━ Grateful to my supervisor A. B. M. Aowlad Hossain Sir for his strong guidance and co-author Tasnimul Hasan for his support throughout this journey. #IEEE #EEG #BCI #xLSTM #DeepLearning #MotorImagery #EmotionRecognition #BCI #Neurotechnology #MachineLearning #RAAICON #BECITHCON #ArtificialIntelligence
-
-
You do not need to be brave to use TiRex - but you should be smart. But first of all you need to understand, what a time series foundation model is doing. Hans Michael Krause asked us the same question, when we started to work with ctrlX. Sepp Hochreiter: "In its chat function, ChatGPT, for example, predicts the responses to questions using so-called tokens. NXAI, on the other hand, predicts the next time series value with its time series foundation model or is able to identify anomalies and emergency situations." Michael understood. With TiRex we speak the language of the machines and processes - forecasting, classifications, embeddings. That makes sense. And than we decided to write an article together for his customers. Btw: Thanks to #xLSTM we run TiRex on Bosch Rexroth ctrlX COREplus X3 (Zync Ultrascale+, 64-bit, 4 × ARM A53) with 2GB RAM on CPU.
-
Disclosure: This is all a coincidence; there is no correlation or causality with current events... 😅 That's not entirely true, because it's the NeurIPS deadline, and we were looking for an interesting paper from last year when we stumbled upon this one. TiRex and forecasting of epidemics 🦠 In comparison, the self-attention mechanism of TimeGPT and the stacked #xLSTM blocks of TiRex enable these models to more effectively learn and predict the complex temporal patterns inherent in epidemic incidence cases. Overall, our empirical evaluation demonstrates that, while statistical and deep learning baselines remain competitive in certain horizons, the architectural innovations of the pre-trained TimeGPT and TiRex models make them especially valuable for policymakers, as they offer robust real-world epidemic forecasts that enable the design of effective intervention strategies. 📝 https://lnkd.in/d4QTJVD3
-
-
We are proud partner of the Industrial AI Podcast 🚀 The Industrial AI Podcast reports weekly on the latest developments in AI for the engineering, robotics, automotive, process and automation industries. The podcast features industrial users, scientists, vendors and startups in the field of Industrial AI. And for us it is the perfect match, because with TiRex we speak the language of the machines and processes (time series) and with our #xLSTM robotics models we really fit to the community that was build around the podcast. You can listen to the podcast on any catcher!
-