Sorin Gheață’s Post

[Product Update] #Alfrada now comes packing with a SOTA on-chip medical vision model from Google called #MedGemma_4b - that has outstanding X-ray and MRI based diagnosis capabilities that you can call in Alfrada by uploading your scans in the form of an image along with a prompt to do medical scanning to either Alfred or Ada. All medical records process exclusively on our private, encrypted NVIDIA #CUDA_Cluster in Frankfurt, this image processing step by MedGemma is by default air-gapped and provides pristine, uncompressed visual diagnosis of thousands of complex medical conditions. #Private_by_Design #Never_Self_Diagnose Hajra Khan Amna Khalid Nabil De Marco Farrukh Aziz Omar Akram Sajjad R. Habibullah Khan Alasdair Redmond Siraj Ahmed Shaikh Aniqa Sandhu Cevat Yerli Sarmad Ali Khan Imran Saqib Shaz Ansari Caner CANAK Michael Birdsall Georgi Gerginov Dr. Anjan Ghosh

The robot is designed to insert hundreds of ultra-fine, flexible threads with thousands of electrodes within microns of targeted neurons while avoiding vasculature and adapting to real-time brain motion.

FPGA Horizons Journal Issue 3 - Proud to see AMD content featured in the latest edition, including insights on: 🔹 Accelerating system‐level verification and 🔹 Building radiation‐tolerant AI platforms for space #AMDEmbedded

"We're building a surgical robot capable of reaching any brain region. The goal: a generalized neural interface to help solve any condition that originates in the brain."

Photon Counting CT -PCCT - provides up to 27 times better resolution than conventional CT, allowing for imaging fine structures in high detail. A game changer technology , a real revolution in cardiac imaging, a faster and reliable way to better definition of cardiac tissue and structure.

To power the next generation of pioneering research, Gemma 4 models are sized to run and fine-tune efficiently on hardware— from laptop GPUs, all the way up to developer workstations and accelerators. By using these highly optimized models, you can fine-tune Gemma 4 to achieve state-of-the-art performance on your specific tasks. We've already seen incredible success with this approach; for instance we worked with Yale University on Cell2Sentence-Scale to discover new pathways for cancer therapy. Learn more about Gemma 4, our most intelligent open models to date: https://goo.gle/4upVyMG

[#arXivaria] Today, in two fresh papers, we can compare blueprints for scalable fault-tolerant architectures using trapped ions and neutral atoms. While the trapped-ion approach relies on mobile qubits that distribute entanglement from GHZ factories, the neutral atom case uses gate teleportation and lattice surgery from nearby logical ancillas. Both methods significantly cut the expected time of a meaningful fault-tolerant simulation (~100 Heisenberg/Fermi-Hubbard spins) to months (ions) or even hours (neutral atoms) using only ~10k physical qubits. It might be worth considering how we might combine fast operation times with qubit mobility to replicate this trapped-ion approach in spin qubits. Neutral atoms: https://lnkd.in/e5uZ-tjn Trapped ions: https://lnkd.in/eeKC6t5h

Day 1 of the AMD Hackathon is live, and I am officially in the solo run for OncoTriage. All stations this is Reaper Eagle One. I have the con The mission is simple: solve the "black box" problem in medical AI. Today’s focus is migrating my EfficientNet-B4 backbone to the AMD Instinct MI300X instances. By implementing MC Dropout for Bayesian Uncertainty Quantification, the system doesn't just identify lung nodules—it quantifies its own confidence. Technical Payload: Hardware: AMD Instinct MI300X (ROCm optimized). Architecture: Bayesian EfficientNet-B4. Objective: Reducing false negatives in chest X-ray diagnostics through probabilistic deep learning. Passive AI clankers are yesterday’s news. It’s time for high-performance, uncertainty-aware clinical intelligence. lablab.ai | AMD Developer #AMD #MI300X #OncoTriage #BayesianDeepLearning #MedicalAI #SoloFounder lablab.ai AMD AMD Developer In the pic below I processed a 563 image batch within less than 3 seconds with surgical accuracy.

Convex cylindrical lenses play a critical role in modern optical systems — focusing light in one dimension while enabling precise beam shaping and line generation. From laser scanning and medical imaging to barcode readers and industrial inspection, these lenses help improve accuracy, efficiency, and optical performance across industries. As photonics and laser applications continue to evolve, high-precision cylindrical optics are becoming more important than ever. #Optics #Photonics #LaserTechnology #OpticalEngineering #CylindricalLens #PrecisionOptics #Innovation