Trustwise steers AI medical education vision

An innovator in AI performance and risk management, the company has collaborated with Health Innovation Kent Surrey Sussex (Health Innovation KSS), Hitachi Digital Services, and Further to co-develop MedAssist GPT. This is a generative AI application for medical education that leverages NVIDIA NIM microservices.

The application equips UK medical students with real-time, personalised evidence-based clinical guidance. Integrating more than 700 National Institute for Health and Care Excellence (NICE) guidelines and NHS Formulary data, MedAssist GPT covers over 30,000 medicines – offering what Trustwise calls ‘unparalleled resources for clinical learning.’

MedAssist GPT is hosted on Google Cloud Platform and leverages the NVIDIA NIM microservice for the Writer Palmyra-Med large language model (LLM). The application will enable UK medical students to engage in clinical scenario simulations, practice safe and cost-effective prescribing, and receive instant feedback.

MedAssist GPT supports over a dozen languages, delivering a personalised learning experience while strictly aligning with national healthcare standards. Trustwise says.

The company claims that MedAssist GPT represents an advancement in healthcare training, overcoming key AI challenges to achieve high safety and reliability standards.

Developers of the application tackled issues like AI output alignment, sensitive data leakage, and hallucinations, ensuring the tool provides accurate, evidence-based medical guidance tailored to regional care protocols.

To anchor students’ learning in real-world practices, MedAssist GPT is said to be meticulously aligned with NICE guidelines, drug formulary data and UK healthcare regulations.

Generative AI speed and performance were prioritised to provide real-time guidance during clinical simulations. The teams at Health Innovation KSS, Trustwise, Hitachi Digital Services, and Further optimised MedAssist GPT to deliver low latency, high-throughput experiences, ensuring seamless operation and accurate, instantaneous feedback.

Balancing these technical advancements with operational efficiency, MedAssist GPT was designed using NVIDIA AI to maintain performance while containing costs, reducing LLM token inefficiencies and streamlining overall AI operations.

In adhering to NHS sustainability goals, the team worked to minimise MedAssist GPT’s carbon footprint, embracing environmentally responsible AI practices as part of the company’s mission to support a sustainable future, Trustwise adds.

The initial success of MedAssist GPT can be attributed to the close collaboration between several key partners. Trustwise contributed its Optimize:ai software, a single API that acts as a trust and optimisation layer, enhancing MedAssist GPT’s safety, alignment and responsiveness.

Optimize:ai minimised token costs and ensured adherence to regional healthcare standards, making MedAssist GPT both effective and efficient.

A member of NVIDIA Inception, a program to advance cutting-edge startups, Trustwise used NVIDIA NIM, part of the NVIDIA AI Enterprise software platform, to enable high-speed inferencing and cross-platform scalability, reducing latency for real-time performance.

Writer’s Palmyra-Med LLM addressed the specific linguistic and knowledge needs of the medical and life sciences fields. Further’s healthcare expertise helped fine-tune AI models and refine the conversational user experience, ensuring the application met practical medical training requirements.

Early results from the beta testing of MedAssist GPT have been promising, Trustwise reveals. In initial beta tests, the application has shown over a 90 per cent improvement in AI output alignment with NICE guidelines, greatly enhancing the accuracy of medical recommendations, the company adds.

It says response times have improved by 30 per cent, providing students with faster feedback during clinical simulations. Operational efficiency has also been boosted, with a 40 per cent reduction in AI token costs.

Notably, MedAssist GPT has also demonstrated a 64 per cent decrease in energy consumption and carbon emissions, supporting NHS sustainability objectives. Trustwise says.

Looking ahead, Health Innovation KSS, part of the national Health Innovation Network established by NHS England, will continue to work closely with medical students at the newly established School of Medicine, within the University of Surrey over the next year to rigorously test and refine MedAssist GPT.

This process will ensure the application’s continued improvement and scalability, ultimately preparing it for a full-scale rollout. The goal is to equip the next generation of healthcare professionals with AI-driven tools that enhance their training and improve patient care outcomes.

“Our vision with MedAssist GPT was to set a new standard in healthcare training, where safety, speed, and sustainability converge,” said Manoj Saxena, CEO and founder of Trustwise.

“The design and development process between the Trustwise, Health Innovation KSS, Hitachi Digital Services, and Further teams was a truly unifying process with brilliant minds in AI development.

“We’re proud to have created a tool that not only empowers students with real-time, evidence-based insights but also aligns with our commitment to trustworthy AI.”

Founded in 2022 by a successful serial entrepreneur and the first general manager of IBM Watson, Trustwise is backed by leading investors and is headquartered in Austin with research labs in Cambridge UK, and New York.

The company is steered by what is regarded as a distinguished group of executives and machine learning experts in trustworthy AI from leading institutions such as the University of Cambridge, the Alan Turing Institute, and NYU.