A team of researchers including @Jintao_Zhang_ investigated one of the most critical challenges in implementing artificial intelligence in healthcare: whether clinicians can reliably identify errors produced by large language models (LLMs). The results show a serious need for an extra line of defense that can co-exist with human decisionmaking. For us at Kōsten, it is a lesson about design. Safe clinical AI needs workflows designed around risk, supported by tools built to show their reasoning and sources, and ongoing monitoring once they’re in use. When the research shows us problems we're here to solve, our ears perk up. Read chief medical officer Gustavo Breitbart's summary below. kostendigital.com/news/clini…

Our new blog post dives into emerging devices that could transform how patients and clinicians manage chronic diseases, like epilepsy. Young and talented researchers at @PennNeurology led by @ZackGoldblum built a system that lets people with epilepsy chat with an AI connected to their monitors. Participants could ask about their seizures, sleep, and brain activity, opening a two-way channel where patient feedback retrained the seizure detector, cutting false alarms by about 80%. It's an early study with few patients but it points to a promising model for patient-AI collaboration that improves the tools that support their care. Read our full breakdown on the blog right here: kostendigital.com/news/patie…

Six-month audit backlogs. Up to 40% of invoices with billing errors every month. A team checking every line by hand and falling behind. Sound familiar? That's where one of our smaller clients started. Wrong prices, wrong quantities, and charges that didn't match the diagnosis were slipping through because there weren't enough hours in the day to catch them with a small team. We designed and built an AI centric auditing platform that now reviews every invoice as it arrives, flags those errors in real time, and fits into the team's existing workflow with HIPAA-aligned security. If your team is stuck doing work a system should handle, let's talk here: kostendigital.com/lets-talk

Researchers at Technische Universität Dresden recently came up with an AI system that diagnoses simulated patients entirely on local servers that kept data private. When the AI gave the same diagnosis across repeated tries, it was usually right, a better sign than its own stated confidence. Our chief medical officer Gustavo Breitbart's takeaway from this incredible research finds that consistency is a practical trust signal for clinical AI. Let it handle the cases it answers the same way every time while routing uncertain cases to physicians. Could this be a glimpse into the future of medicine? Our team at Kōsten Digital covers the latest in healthcare AI research, not as an outsider, but an important chain in the system. Read our summary of the research here: kostendigital.com/news/simpl… @jnkath @katherlab @Dykex6 @gwoelflein @Junhao_Liang

The majority of healthcare still runs on manual processes and disconnected data. Information sits in separate systems, making teams re-enter the same details by hand and burying valuable insights. These outdated corners of the industry are where AI can make some of the biggest efficiency gains which are hard to argue against. Kōsten Digital was created to do just that: design, build, and operate custom and exclusive solutions for health systems, biotech, and pharma. That includes bringing scattered data together and reducing repetitive manual work that can derail billion-dollar projects. Check out our revamped corporate deck on our homepage, where you'll find our past successes and what makes us different. Let's chat today: kostendigital.com/lets-talk

What if every cell could be compared on a single map? Our latest blog looks at a new AI foundation model from @Stanford's brilliant team of researchers @YanayRosen, @yusufroohani, @ayushh_agrawal, @StephenQuake, @jure and others that learns to represent cells across tissues, experiments and species. Trained on more than 36 million cells, it built an atlas of over 1,000 cell types and can even map cells from species it has never seen. For biomedical research, a shared map like this could help scientists compare new samples against millions of known cells, spot unusual cell states and generate new hypotheses about disease. It's a meaningful step forward in advanced medicine, an area we prepare for the AI era. Read the full summary here: kostendigital.com/news/unive…

A conversation is just the start. And one with no end when you partner with our team at Kōsten. Most technology vendors stop at "build" while we stay online to monitor and adjust what we implement for our clients over the long run. Having full-time physicians and scientists on staff means our custom solution designs begin with how clinicians actually work, because if isn't a fit for your team, why do it? Chat with us today to get started: kostendigital.com/lets-talk

AI could function as an additional surveillance layer for immunotherapy clinical trials, an essential component of patient safety and regulatory drug evaluation. In this week's blog post, we summarize the latest from @UCSF researchers @danielkwonmd and others who found that AI was able to identify adverse events about as well as individual physicians, and even caught some they missed. Full blog right here: kostendigital.com/news/what-…

Kōsten Digital's edge is its people. From client onboarding to data analytics to applying AI-first solutions that make the difference, our team of technology and healthcare experts are who you want in the dirt to make things easy. Each individual is at the top of their game with past experience in clinical trials, developing historical cancer treatments, and solving billion-dollar problems in a matter of weeks. It's who you want in your toolbox at all times. The best part? You can start a conversation with any of them today, right here: kostendigital.com/lets-talk

This week, we'd like to highlight new AI x Sepsis research from the talented @ABoussina and Kimberly Quintero out of @UCSDHealth. Sepsis is one of the leading causes of hospital deaths, and this team asked themselves: could generative AI move beyond retrospective analysis of electronic health records and become part of a real-time clinical quality improvement system? In other words, could physicians improve sepsis care faster by getting feedback in the moment? The results showed that adherence to sepsis standards improved, and the AI's reviews closely matched those of expert human reviewers. The idea could reshape how hospitals get better: continuous, timely feedback drawn from everyday care and beyond. Read our full summary here: kostendigital.com/news/impro…

Human-dependent invoice reviews are slow, inconsistent, and error-prone. Many contain billing discrepancies which cause major financial losses over time. Add to that six month + backlogs and no path to scalability. One of our clients needed a solution. So we built them an exclusive AI-powered auditing platform to change that — AI agents + LLMs that validate every invoice automatically, cross-checking pricing, quantities, and diagnosis relevance in real time. Results: ✅ AI agents for automated discrepancy detection ✅ Seamless integration with existing insurer workflows ✅ Structured dashboards for actionable insights ✅ Local anonymization and encryption ✅ Compliance-first, HIPAA-aligned design Proud of what our team built here. This is the kind of digital solutions work that actually moves the needle in healthcare. How can we help? Let's talk: kostendigital.com/lets-talk

It's easy to get pulled into the 'doomsday' narratives around AI, especially in recent weeks. But according to academics, patients, and medical experts, the biggest positives from this groundbreaking technology has been medical breakthroughs. In as little as a year, terminal illnesses have become easier to treat, extending life for people around the world. AI is helping clinicians catch diseases earlier, personalize treatment plans, and find options where none existed before. At a time when healthcare organizations are feeling the pressure that goes with this, it's never been more important find an efficiency edge. At Kōsten, we build the infrastructure that makes this possible — designing and operating the digital data systems that turn data into tools clinicians can actually use. The pace of progress in healthcare AI is accelerating. We're glad to be part of making it usable. Let's work together: kostendigital.com/lets-talk

AI is incredible at diagnosing and managing cardiovascular disease, but patient outcomes are the true measure of success. A new systematic review from seasoned researchers across London, Singapore and Malaysia offer an encouraging—but nuanced—answer. Read the full blog to explore what the evidence tells us about where AI is making a difference, where the gaps remain, and what the next generation of clinical AI research needs to prove. 👉 Read the full post here: kostendigital.com/news/beyon… @DrFuSiongNg @ejosipcar

There's a lot of AI talk going around. But what separate us from most is that AI is not at the core of what we do. We are 'efficiency edge' seekers that leverage our team's subject matter + technology expertise to give companies in the healthcare space the extra inch over their competitors and the extra mile toward their goals. Our results are anything but artificial. Let's talk today to get started: kostendigital.com/lets-talk

A new research paper out of Harvard University shows a way to extract messy clinical data without ever moving sensitive patient info outside the hospital: a small but mighty LLM that puts privacy first. Tested across three cancer cohorts, it matched a model nearly five times its size — and turned 2.5 hours of automated extraction into what normally takes two weeks of manual chart review, with nearly identical accuracy. Check out our full summary here: kostendigital.com/news/what-…