NoHarm.ai prevents medication errors and saves lives using AWS AI

The Brazilian Unified Health System (SUS) is the sole public health provider for about 75 percent of the population, making public hospitals critical access points for local communities. However, resource and staffing constraints present constant challenges to patient care: A single clinical pharmacist in a public hospital, for example, might be responsible for hundreds of patients, while remote Amazon River clinics often lack critical infrastructure such as electricity and internet.

Documented medication-related errors, ranging from prescribing and dispensing to actual administration, have led to intensive care unit admissions and deaths. These errors are a leading and preventable cause of patient harm, with the consequences falling hardest on underserved populations facing geographic isolation or socioeconomic constraints. “Errors occur even in the most critical clinical areas, such as neonatal and pediatric care,” says Henrique Dias, cofounder of NoHarm.

As a clinical pharmacist at a large public hospital in Porto Alegre, Brazil, NoHarm cofounder Ana Helena Ulbrich lived this reality firsthand. She and five colleagues reviewed nearly 1,000 prescriptions a day; the safe clinical workload is generally considered 25 patients per pharmacist per day. In one case, a patient received 10 times the prescribed dose of methylene blue, a therapeutic drug that can be toxic at high levels. The error was detected only because the patient was in the intensive care unit, where constant monitoring caught it in time. “There are errors we don’t find out about,” says Ulbrich. “If we knew, we wouldn’t let them happen.”

NoHarm’s AI-powered solution is designed to catch medication-related errors that could otherwise go undetected, such as the misdosage of morphine for a newborn in neonatal intensive care. The solution advances health equity by protecting the most vulnerable patients in settings where limited resources and staffing constraints make consistent prescription review a challenge.

NoHarm selected AWS as its infrastructure provider for flexibility, cost-effectiveness, and compliance with the Brazilian Data Protection Law, which requires that health data remain within the country. Because the solution is provided free of charge to public hospitals nationwide, keeping infrastructure costs low is essential. By building on AWS, NoHarm has maintained a reliable, sustainable, and affordable infrastructure that supports its no-cost model for public hospitals.

With support from the AWS Health Equity Initiative to offset technology costs, NoHarm combined rules-based safety controls with advanced AI capabilities to create a solution for evaluating prescriptions as they are created or modified. It performs precision extraction of key data points by using named-entity recognition to highlight critical information within clinical notes. Then, it generates annotated discharge summaries through contextual summarization with large language models in Amazon Bedrock, a service for building generative AI applications.

NoHarm’s solution also cross-references prescriptions against laboratory results to identify potential dosing issues, therapeutic duplication, drug interactions, and contraindications, a term referring to specific situations, symptoms, or medical conditions that make a treatment, medication, or procedure inadvisable because it could be harmful to the patient.

A scoring system in NoHarm’s solution helps pharmacists prioritize high-risk cases, reducing missed alerts that can lead to preventable harm. Human-in-the-loop oversight and clinical pharmacist feedback refine model accuracy over time, boosting trust in AI-assisted clinical decisions. A clear audit trail in the solution also supports regulatory compliance.

By 2025, NoHarm had deployed its solution across 182 healthcare units in 17 Brazilian states. The platform explicitly tracks 137 public-sector hospitals, comprising 77 SUS public hospitals and 60 mixed public-private hospitals, and monitors 23,613 public-system beds—roughly 59.5 percent of SUS’s total bed capacity. NoHarm’s solution has positively impacted more than 4 million patients through AI-driven prescription review, analyzes more than 5 million prescriptions monthly, and processes more than 400,000 clinical notes daily.

Clinical pharmacists have used NoHarm’s solution to correct millions of flagged errors before they could cause harm, and hospitals report up to 800 percent faster prescription analysis. One hospital reduced costs by R$2 million after 2 years, redirecting those resources into essential medications, infrastructure, and sustaining public hospital beds.

A peer-reviewed clinical study at a teaching hospital serving about 1.5 million people also found that the solution raised the share of prescriptions reviewed from 0.6 to 49 percent. At the same time, errors dropped by almost 98 percent in 6 months. Through the Juntos Pela Saúde (Together for Health) initiative, NoHarm collaborates with the Brazilian Development Bank to extend its solution to 20 municipalities, including remote Amazon River villages, supplying solar panels, computers, and satellite internet to clinicians.

NoHarm’s AI-powered solution supports professionals working within a system built on equitable access to health—regardless of a patient’s income, location, or resources. “We want NoHarm to flag every type of risk and to serve as an increasingly strong safeguard protecting patients from harm,” says Dias. “Using AWS AI, our goal is that no patient ever experiences a problem because of a medication error.”