Meet Russell-GPT: NUHS’ homegrown AI transforming medical coding

Published on 9 March 2026

Built in-house, this artificial intelligence system is taking the slog out of medical coding – and changing how hospital teams get things done.

At a glance

• Medical coding is essential but painstaking and is often a hidden choke point in hospital workflows.
• Russell-GPT, NUHS’ in-house AI, speeds up the work by extracting and summarising key patient information.
• Early results show quicker turnaround and better accuracy, with plans to scale system-wide by 2030.

Chances are, when you hear “coding”, you picture programmers, not patients. But in healthcare, medical coding is what keeps the system ticking – turning patient health information such as diagnoses, treatments and tests into standardised alphanumeric codes used for billing, insurance and Medisave claims, and government subsidies.

Without accurate codes, payments stall, subsidies get miscalculated, and the hospital’s financial machinery grinds to a halt.

However, medical coding is often laborious and painstaking. Coders wade through mountains of clinical documents, deciphering notes and results before they can even begin assigning the right codes. It is slow, tiring work, and one of the least visible choke points in hospital workflows.

Enter Russell-GPT (RGPT) – an artificial intelligence (AI) assistant at the National University Health System (NUHS) that speeds up the process and improves coding accuracy. RGPT is just one of several practice-changing innovations at NUHS that transform how work is done and how care is delivered.

Taking the grind out of coding

“Before RGPT, coders had to go through all the clinical documents within EPIC (NUHS’ electronic record system) before they could start coding,” said Ms Jennie Sun, Head, Group Medical Records Office, NUHS. “Now, with the prompts supplied by the coders, the AI extracts the key medical information for them.”

Those documents include everything from patient consultations and lab results to surgery notes and discharge summaries. “Coders no longer need to comb through all of that,” said Ms Sun. “RGPT summarises it all, which saves a lot of time.”

To be clear, the AI does not do the coding. It gathers and condenses the information so that the coders can focus on applying the actual codes. The result: fewer bottlenecks, faster processing and sharper accuracy. “It also enables coders to spend more time reviewing more complicated cases,” Ms Sun added.

Currently, RGPT achieves about 90 per cent accuracy. The remaining 10 per cent still requires manual review. “Human oversight is still needed because of coding rules and what is lacking in the AI,” said Ms Sun. “From the pilot run and our pre-investigations, we know there is a certain percentage that the AI will not be able to handle. Those still require manual checks and investigation before final coding.”

A process of training and learning

Russell-GPT is still in its first phase, a period of trial and error during which the AI learns what the coders want. They, in turn, learn how best to teach it – questioning the output, spotting mistakes, investigating further, and refining their prompts. “It is a learning journey,” said Ms Sun. “We have to teach the AI to be more accurate.”

After a pilot run in October 2024, the system went live in August 2025. “We are starting really small,” Ms Sun said. “We are currently using RGPT only for the very simplest cases, covering about 6 to 8 per cent of our total workload.”

The first phase will continue into 2027–2028. Over time, the AI’s role may expand beyond coding support. “Clinicians have approached me about using it to improve their clinical documentation further down the road,” said Ms Sun. “There’s definitely potential to build something that can alert doctors to discrepancies, inaccuracies or signs that something is amiss.”

The long-term goal is ambitious, with efforts focused on completing the rollout by 2030. Ms Sun said: “By that time, our hope is that all three hospitals – National University Hospital (NUH), Ng Teng Fong General Hospital (NTFGH) and Alexandra Hospital (AH) – will be using RGPT and achieving a 25 per cent time savings in coding.”

In consultation with Ms Jennie Sun, Head, Group Medical Records Office, NUHS, and Dr Andrew Makmur, Group Chief Technology Officer, NUHS.