MManfye Goh / Clinical Intelligence
Malaysia / Founder, Wistify

Manfye Goh —
Clinical insight.
Working systems.

I’m a pharmacist, software engineer and founder of Wistify. I turn the friction of everyday care into software, applied AI and tools that help professionals serve people better.

Explore selected systems
Open fieldbook / 04

The work behind the profile.

Public work and experiments across healthcare, AI, and software systems.

01 / Care deliveryProduct & engineering practice

medcare

Start with the professional’s workflow.

Healthcare software has to fit the work around a consultation, not just capture the consultation itself.

My work

I build pharmacy and care-workflow tools, connecting clinical context with practical software. My public engineering notes also document incremental modernization: introducing a gateway and moving modules while old and new systems coexist.

Engineering sketch / incremental migration
Existing workflowsGateway layerModules evolve

Old and new systems coexist while the transition happens.

Read the engineering field note ↗︎
medcare: design notes

A useful system earns its place in everyday work. Reliability, migration and a clear next action matter as much as a new feature.

02 / Clinical AIResearch / 2025 preprint

MedGemma & multimodal RAG

Connect images, language and clinical references.

Clinical information arrives in more than one format. Text retrieval alone cannot describe everything in a medical image.

My work

I co-authored work on fine-tuning MedGemma for clinical captioning in a multimodal retrieval workflow over Malaysian clinical practice guidelines.

Research sketch / multimodal retrieval
Medical imageClinical captionRelevant guidance

A conceptual research workflow, not an autonomous diagnostic tool.

Read the preprint ↗︎
MedGemma & multimodal RAG: design notes

Evaluation belongs beside the demonstration. Research results need their dataset, scope and limitations, not a blanket claim of clinical readiness.

03 / Regulatory workflowsDevelopment / public project account

RegularOS

Make evidence easier to inspect.

Pharmaceutical regulatory work brings together evidence spread across documents and formats.

My work

My public work explores AI-supported regulatory affairs: organizing pharmaceutical evidence into traceable, human-reviewable workflows.

Read the project post ↗︎
RegularOS: design notes

The professional should be able to see where an answer came from and decide what happens next.

04 / The experimental benchOpen-source experiment

Countenance

Explore intelligence on the device.

What can a browser learn from a sequence of visual signals while keeping processing close to the user?

My work

Countenance is my open-source facial-expression extension experiment. It brings computer vision and lightweight temporal modeling into a browser-based project.

Countenance demonstration interface showing locally stored session-history labels and a chart
Public repository demo: local session-history interface. These model labels are not evidence of a person’s internal emotional state.
Explore the public repository ↗︎
Countenance: design notes

An expression classifier is an experiment in visual signals, not a reliable reading of someone’s feelings or intentions.

Products, research and experiments have different standards of evidence. Each entry links to its public source; project descriptions are not claims of clinical effectiveness.

Research output / 04

Publications

Peer-reviewed and preprint work spanning clinical language models, medical imaging, and health-product recommendation systems.

01
Preprint / Medical imaging AI

Fine-Tuning MedGemma for Clinical Captioning to Enhance Multimodal RAG over Malaysia CPGs

How specialized image captions can support retrieval over Malaysian clinical guidelines.

Lee Qi Zun · Mohamad Zulhilmi Bin Abdul Halim · Goh Man Fye

arXiv
202502
Preprint / Ophthalmic AI

FundaQ-8: A Clinically-Inspired Scoring Framework for Automated Fundus Image Quality Assessment

Eight clinically inspired quality criteria make fundus-image assessment more explicit.

Lee Qi Zun · Oscar Wong Jin Hao · Nor Anita Binti Che Omar · Zalifa Zakiah Binti Asnir · Mohamad Sabri bin Sinal Zainal · Goh Man Fye

arXiv
202503
Conference paper / Clinical NLP

Enhancing Medical Summarization with Parameter Efficient Fine Tuning on Local CPUs

Exploring parameter-efficient adaptation for medical summarization on local hardware.

Shamus Sim Zi Yang · Goh Man Fye · Wei Chung Yap · Diong Zi Yu

IEEE ICECCE · DOI 10.1109/ICECCE63537.2024.10823619
202404
Journal article / Recommender systems

A hybrid recommender system for health supplement e-commerce based on customer data implicit ratings

Using implicit customer feedback to study health-supplement recommendations.

Pantea Keikhosrokiani · Goh Man Fye

Multimedia Tools and Applications 83 · DOI 10.1007/s11042-023-17321-6
2024
Manfye Goh, from his public GitHub profile
Manfye Goh
Public GitHub profile
Profile signal

A pharmacist who learned to build the tools healthcare was missing.

I’m Manfye, a pharmacist who builds software. I founded Wistify to turn clinical understanding into practical tools for the people delivering care. My work crosses pharmacy workflows, applied AI, data science and the engineering needed to keep a product useful.

My path includes pharmacy IT and informatics work in Malaysia’s Ministry of Health, a Master’s in Data Science & Analytics, and AI leadership at Qmed Asia. Today, Wistify is where I bring that clinical and technical experience together, working with a team to build and improve medcare.

I care about the space between an interesting demonstration and something a professional can actually use: the missing context, the awkward handover, the repeated task, and the follow-up that needs to happen.

Explore the public workGitHub profile ↗︎Hugging Face profile ↗︎LinkedIn profile ↗︎
Selected public milestones

The thread runs
through pharmacy.

Different tools, one continuing interest: making clinical knowledge easier to use. This is a trail of public work, not a complete employment timeline.

  1. Clinical references, made practical

    Early writing under my “manfye” byline documents pharmacy tools, including MediQuest and a Malaysian dengue-guideline app.

    Explore the early archive ↗︎

  2. Sharing AI with healthcare learners

    AIMST documented my AI-in-healthcare workshop, bringing an informatics pharmacist’s perspective to a changing technical field.

    Read the institutional account, page 33 ↗︎

  3. Research with the Qmed team

    Co-authored FundaQ-8 and MedGemma clinical-captioning research. The askCPG team’s multimodal work also appears in Google’s Health AI Developer Foundations showcase.

    Explore the askCPG showcase ↗︎

  4. Building through Wistify

    Working on medcare, publishing experiments, and keeping the clinical problem connected to the engineering decisions.

    Follow Wistify’s work ↗︎
Beyond the usual interface

Sometimes the prototype
leaves the screen.

I also explore spatial and hands-on learning. One public prototype uses Meta Quest for a simulated pharmacist–patient interaction, combining spoken counselling and medicine-handling tasks. It is a demonstration of an idea, not evidence of educational effectiveness.

Watch the pharmacy VR prototype ↗︎
My working method

From clinical friction
to something useful.

My pharmacy background shapes what I notice. Engineering lets me act on it. Evaluation helps me decide what deserves to stay.

  1. 01 / Observe

    Observe the handover.

    A feature starts with who needs the information, what is missing, and what action follows.

  2. 02 / Build

    Build around the decision.

    Sometimes the answer is a better workflow. Sometimes it is a model. The problem decides.

  3. 03 / Evaluate

    Expose the evidence.

    Make sources, uncertainty and evaluation visible enough for a professional to question them.

  4. 04 / Improve

    Stay with the system.

    Maintenance, migration and feedback are part of the product, not an afterthought.

Writing & teaching

Make the thinking visible.

Building is only part of the work. Explaining decisions helps other people test, adapt and challenge them.

Perspective

Medical AI sovereignty

Why local clinical context and clinician participation matter.

Read the essay ↗︎Technical field notes

From models to interfaces

Writing on medical imaging, data, retrieval and building useful software.

Explore the writing ↗︎Speaking / 2025 archive

Clinical practice meets AI

Professional background and speaking at Mi-PHARM. Historical role descriptions reflect that event.

View the speaker profile ↗︎
At the intersection

Let’s work on a real problem.

I’m interested in conversations where clinical knowledge and technical execution need to meet: healthcare products, applied AI research, and practical learning for professionals.

Discuss a project on LinkedIn ↗︎
Clinical workflow × ProductResearch × EngineeringProfessional learning × Practice