Neurotechnology Innovations

No gradients. No backprop. Just projections — a fundamentally different, mathematically grounded approach to neural network training that scales. Joint work with Manish Krishan Lal, Stefanie Jegelka and Suvrit Sra. 📄 https://lnkd.in/eajcfeH3 Here’s how it works 🧵 • We reformulate training as a 𝗳𝗲𝗮𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆 𝗽𝗿𝗼𝗯𝗹𝗲𝗺, not loss minimization. • Each neuron and data point add a constraint. • We then project onto the constraint sets, finding a point that satisfies all constraints = a trained model. Why this is cool: 1. projections are 𝗰𝗵𝗲𝗮𝗽; roughly the cost of a forward pass 2. they can be computed independently across neurons and data points -> 𝗶𝗻𝗵𝗲𝗿𝗲𝗻𝘁 𝗽𝗮𝗿𝗮𝗹𝗹𝗲𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 3. natural support for 𝗻𝗼𝗻-𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁𝗶𝗮𝗯𝗹𝗲 components and 𝗵𝗮𝗿𝗱 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀 We built a whole framework for this: 𝗣𝗝𝗔𝗫 • Think autodiff for projections. • Built on JAX, it inherits hardware acceleration & JIT, with a familiar interface. • We trained MLPs, CNNs, and RNNs with PJAX. • 🔗 https://lnkd.in/ea4pc-SG Looking forward to the community's response! The approach has potential beyond standard training — particularly for tasks with non-differentiable components or local constraints, like 𝗽𝗿𝘂𝗻𝗶𝗻𝗴, 𝗾𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻, and 𝘀𝗽𝗮𝗿𝘀𝗲 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴.

There’s an emerging biotech market which I expect we’ll be hearing more and more about over the next 5 years - and we’ve got Australian startups playing a role. When cells talk to each other, they don’t just signal with proteins and hormones. They also pass tiny packages called exosomes - which could increasingly be the future of therapeutics, diagnostics, and cosmetics. 🧪 Here’s the science: Exosomes are tiny vesicles (30–150 nm) that cells release into their environment. They’re essentially lipid-wrapped bubbles carrying proteins, RNAs, lipids, and metabolites, which mirror the parent cell. You can look at exosomes to get a good understanding of the cell they originated from. 🚀 Here’s why they’re useful and where (I expect) the market is going: Diagnostics: Since exosomes are naturally present in most bodily fluids, so they’re a non-invasive way to get a read on diseases like cancer, cardiovascular disease, and neurodegeneration. Therapeutics: Exosomes naturally deliver “packages” of information from cell to cell; sounds like an ideal vehicle for drug delivery. For example, RNA therapies could be packaged and sent with the body’s own wrapping paper plus delivery mechanisms, making uptake super effective and free from side-effects. Cosmetics: Exosomes are being used for skin rejuvenation by reducing inflammation and boosting collagen/elastin production. They can also be used to stimulate hair growth. However, this is an area where current regulation varies; in what instances should exosomes be regulated as biologics? Plant-derived exosomes are sometimes used to avoid regulatory measures, but their biologic effectiveness is under debate. ⁉️ The biggest challenges currently seem to be around scale and standardisation. Isolating pure, reproducible exosomes is still tricky and those regulatory frameworks are only just becoming established. Still, I wouldn’t be surprised if “exosome” becomes a general-public buzzword in the next 5 years, as noninvasive diagnostics and off-the-shelf regenerative treatments begin to make their way into the mass markets. 👩🔬 Here are the Australian companies working in this space (that I’m aware of - there may be others!) Exopharm Ltd is building the platforms for large-scale manufacturing of exosomes and exosome-based therapeutics. VivaZome Therapeutics is developing exosome therapies for life-threatening disorders for which current treatments are ineffective Exosome Biosciences has pioneered the world's first-in-human clinical trial of an extracellular vesicle therapy for a specific condition. This company is a spin-out to commercialise IP developed at Monash Health, Monash University and the Hudson Institute of Medical Research 🌎 Keen to hear about other companies world-wide working on exosomes for an end-goal you think is particularly cool!

Neurotheranostics: The New Frontier in Nuclear Medicine In the ever-evolving landscape of radiology, few fields hold as much promise and urgency as Neurotheranostics. While Theranostics has already revolutionized care for certain cancers, especially in prostate and neuroendocrine tumors, we are now standing on the edge of a new frontier: the brain. As a radiologist working at GE HealthCare, I’ve seen firsthand how the fusion of diagnostics and therapy can transform lives. Now imagine applying that same precision to neurodegenerative diseases like Alzheimer’s, Parkinson’s, or glioblastoma - conditions where time is brain, and therapeutic windows are narrow. That’s the vision of Neurotheranostics. So what exactly is it? Neurotheranostics merges molecular imaging with targeted therapy in the central nervous system. By using highly specific radiotracers, we can see pathology at a molecular level long before clinical symptoms emerge. Even more exciting: we can potentially treat it with radioligand therapies that home in on those same targets. From amyloid- and tau-targeted agents to radiolabeled molecules that bind to glial activation or neuroinflammation, the possibilities are vast. This could fundamentally shift how we manage neurodegenerative disease - moving from reactive care to proactive intervention. But the science alone isn’t enough. We need the right infrastructure, training, and technology to make Neurotheranostics a reality. That’s why at GE HealthCare, we are doubling down - investing in radiopharmaceutical innovation, PET/CT and SPECT/CT imaging, and AI-enabled workflows that connect the dots between diagnosis, decision-making, and delivery of care. Why now? Because the need is urgent. Neurological disorders are the leading cause of disability worldwide. Traditional therapies often arrive too late or target too broadly. Neurotheranostics offers a more personal, precise approach treating the disease, not just the symptoms. Just as we once reimagined cancer care through Theranostics, we must now reimagine brain care with the same boldness. This is not science fiction. It’s science with vision. The path ahead will require collaboration between radiologists, neurologists, nuclear medicine specialists, pharma partners, and policy makers. But the potential payoff is immense: earlier diagnosis, tailored treatments, and perhaps one day prevention. Let’s push the boundaries of what’s possible. Let’s bring Theranostics to the brain. Let’s create a world here healthcare has no limits. The next frontier is here - and it’s called Neurotheranostics. #gehealthcare #radiology #nuclearmedicine #theranostics #digitalhealth

Collaborative innovation combining AI with neuropsychology is proving to be transformative. Six research clusters show specific value and potential: 🌱 Neuroscience and Mental Health: Understanding mental health through neuroimaging and machine learning enables earlier, more precise interventions for conditions like ADHD and depression. By examining correlations in brain function, this research helps identify key markers for cognitive impairments, aiding in early diagnosis and personalized treatment plans. 🔍 Computational Modeling: Computational models simulate decision-making and cognitive markers, which are crucial for neurological conditions like epilepsy. Machine learning applied to seizure detection, for instance, offers a potential breakthrough in predicting and managing epilepsy, helping patients gain better control and care. 🧠 Cognitive Neuroscience: Studies of cognitive decline and neurodegenerative diseases, such as Alzheimer’s, benefit from reinforcement learning models that reveal patterns in brain degeneration. These insights are essential for developing strategies to slow disease progression, offering hope for more effective interventions. 💡 Cognitive Neurology and Neuropsychology: Examining cognitive functions through neuroimaging and machine learning provides deeper insights into disorders like aphasia and neurocognitive deficits. By mapping brain functions and assessing structural changes, these studies advance our understanding of how specific neurological impairments affect behavior and cognition. 💗 Neuropsychological Features: Machine learning models predict mental health outcomes and cognitive declines by analyzing attention and processing speed. This focus on prediction and prevention, especially for conditions like cardiovascular disease impacting cognition, enables proactive care and lifestyle adjustments to mitigate risks. ⚙️ Neurodegenerative Conditions: AI-based predictive models for neurodegenerative diseases like Parkinson’s allow for early, more accurate diagnoses. By analyzing markers in social cognition and emotional processing, this cluster supports personalized interventions, helping to maintain patient quality of life and reduce care burdens. This is only the beginning. This field is absolutely ripe for rapid advance and massive real-world value.

What if a model could dynamically adapt its compute budget at inference time? Today, we face a rigid trade-off in large language models: either maintain multiple specialised models, or rely on approaches that don’t scale well to modern foundation models. In new work led by my PhD student Paulius Rauba, accepted at #ICLR2026, we introduce Nested Subspace Networks (NSNs) - a way to move beyond this limitation. Rather than designing separate models for different compute budgets, NSNs allow a single model to operate across a continuous spectrum of efficiency and performance. This means we can adapt compute dynamically, without sacrificing robustness or requiring retraining. What is particularly exciting is that this approach enables: - Significant reductions in inference cost - Strong performance even under constrained compute - Smooth adaptability, even in settings not seen during training More broadly, this is part of a shift towards adaptive, resource-aware AI systems - models that can respond intelligently to real-world constraints, rather than operating under fixed assumptions. I’m incredibly proud of this work and of Paulius for pushing this forward. Looking forward to sharing more and connecting at ICLR! Link to paper: https://lnkd.in/eAjZ2R4m

This could be a watershed moment for AI as the 'Deep Learning' era may be evolving into something new. For the last decade, the researchers and engineers have focused on enhancing AI by stacking more layers, which characterizes, the Deep Neural Networks. But a seminal new paper from Google Research for NeurIPS 2025 exposes a fundamental flaw in this approach, these models are static! Once trained, modern models are frozen in time, experiencing a form of 'anterograde amnesia' where they cannot learn from the present without forgetting the past. The paper titled 'Nested Learning: The Illusion of Deep Learning Architectures' by Ali Behrouz, Meisam Razaviyayn, Peiling Zhong, and Vahab Mirrokni proposes a paradigm shift:- Nested Learning (NL). Instead of merely stacking layers, NL reimagines models as a system of 'nested optimization problems', each operating at its own speed. Inspired by human brain waves, where high-frequency neurons manage the immediate present and low-frequency oscillations consolidate long-term memory, this approach unlocks the potential for true continual learning. Additionally, the authors introduced HOPE, a new architecture based on this paradigm. HOPE demonstrates superior performance, surpassing Transformers, RetNet, and Titans in language modeling and reasoning tasks. This could serve as the blueprint for the next generation of AI. Blog - https://lnkd.in/dQ_vermU Paper - https://lnkd.in/di8wnF7r #ArtificialIntelligence #MachineLearning #GoogleResearch #NestedLearning #ContinualLearning #AI

Kolmogorov-Arnold Networks as an alternative to traditional Neural Networks! Researchers from MIT, Caltech, and Northeastern have introduced a new type of neural network architecture known as Kolmogorov-Arnold Networks (KANs), which presents a significant challenge to the traditional use of Multi-Layer Perceptrons (MLPs). KANs offer a novel approach to neural network architecture inspired by the Kolmogorov-Arnold representation theorem. This theorem essentially states that any multivariate continuous function can be represented as a composition of univariate functions and the addition operation. Translating this into neural network design, KANs uniquely place adaptable activation functions on the connections or edges between nodes rather than using standard fixed activation functions at the nodes themselves. This flexibility allows KANs to potentially model complex relationships and patterns more effectively, as they can tailor the transformation at each connection to better suit the specific data and task at hand, diverging from traditional networks where the choice of activation function at each layer is static and uniform across the network. In terms of accuracy, much smaller KANs can achieve comparable or better performance than larger MLPs on tasks such as data fitting and PDE solving. Moreover, KANs demonstrate faster neural scaling laws, meaning their performance improves more rapidly with increased model size compared to MLPs. KANs also excel in interpretability. They can be intuitively visualized and allow for easy interaction with human users. In case studies from knot theory and physics, KANs served as interactive "collaborators" to help scientists rediscover known mathematical and physical laws, showcasing their potential for scientific discovery. KANs could potentially serve as a foundation model for AI+Science applications and open opportunities to improve today's deep learning models that heavily rely on MLPs. Read the full paper for more details: https://lnkd.in/erEF6HbT :)

AI is getting closer to accessing the one thing we’ve always considered private: your thoughts. Recent advances in neuro-AI can now identify whether a person recognizes specific information using EEG signals. A 2025 study using deep-learning reached 86.7% accuracy in detecting recognition through the P300 brain wave: a response triggered before conscious awareness. Meanwhile, some jurisdictions are already experimenting with this technology. 🇮🇳 India has used brain-mapping techniques in hundreds of criminal investigations, showing just how quickly neuroscience can enter real-world decision systems. But the implications go beyond law enforcement. AI models can now (fMRI + diffusion models): Reconstruct visual experiences directly from brain activity ✔️ Models that reconstruct what you’re seeing — in near real-time — based solely on your brain activity (Think: AI generating the images your eyes are looking at.) Decode unspoken language in early experimental settings ✔️ Models that reconstruct the words you’re thinking, even if you never speak A 2023–2024 wave of studies using fMRI + LLMs demonstrated the ability to decode semantic meaning of inner speech—turning thoughts into text-like outputs. This raises critical questions for business leaders, policymakers, and innovators: How do we prepare for a world where cognitive data becomes a new category of sensitive information? What safeguards, standards, and governance frameworks will protect mental privacy as neuro-AI scales? The technology is advancing faster than the regulations around it and the organisations that understand this early will be better positioned to navigate what comes next. #AI #Neuroscience #Innovation #Leadership #Ethics #FutureOfWork Reference: Kim, S., Cheon, J., Kim, T., Kim, S. C., & Im, C.-H. (2025). Improving electroencephalogram-based deception detection in concealed information test under low stimulus heterogeneity. arXiv. https://lnkd.in/dyVqBbG3 Takagi & Nishimoto (2022). High-resolution image reconstruction with latent diffusion models from human brain activity. BioRxiv. https://lnkd.in/dfc32mS7 Tang, J., LeBel, A., Jain, S. et al. Semantic reconstruction of continuous language from non-invasive brain recordings. Nat Neurosci 26, 858–866 (2023). https://lnkd.in/dnQxcS_d

Every second counts in a stroke. When blood flow to the brain is blocked or a vessel ruptures, millions of neurons are lost each minute. The difference between full recovery and lifelong disability often comes down to speed, accuracy, and access to the right treatment. Symptoms can appear suddenly: facial droop, arm weakness, slurred speech, loss of balance, or vision changes. These are moments of crisis where rapid recognition and immediate medical attention save lives. Despite global awareness campaigns, many patients arrive too late for the most effective interventions like clot busting drugs or thrombectomy. This is where artificial intelligence can make a profound difference. 1. Early Detection Algorithms trained on millions of CT and MRI scans can detect subtle changes in brain tissue faster than the human eye. This can alert clinicians immediately, even in hospitals without a full-time neuroradiologist. 2. Triage and Workflow Optimization AI systems can prioritize cases, send automatic alerts, and ensure that stroke teams are activated the moment a scan is uploaded. This reduces the “door-to-needle” time and helps align every step of care. 3. Predictive Analytics By analyzing patient history, vital signs, and lab results, AI can identify those at highest risk before a stroke occurs. This opens the door to prevention strategies and early interventions. 4. Telemedicine Integration AI-powered stroke networks can extend expert care to rural and underserved regions. A patient in a small town can receive the same level of diagnostic precision as one in a major academic hospital. 5. Rehabilitation Support After a stroke, recovery is a marathon. AI-driven rehabilitation tools, including virtual reality and motion tracking, can personalize therapy and track progress, improving outcomes over time. The goal is clear: no patient should suffer preventable disability because the system was too slow to act. With AI as a partner, the chain of survival and recovery can become stronger, faster, and more human-centered. Follow Zain Khalpey, MD, PhD, FACS for more on Ai & Healthcare. Image ref : Mayo Clinic #Stroke #HealthcareInnovation #AI #DigitalHealth #Neurology #StrokeAwareness #HealthTech #AIinMedicine #EmergencyMedicine #PreventiveHealth #BrainHealth #StrokeRecovery #Telemedicine #ClinicalAI #MedicalImaging #FutureOfHealthcare #PatientCare #HealthcareEquity #InnovationInHealth #StrokeSurvivor