Line 01
Governed AI platforms
Custom governed AI systems built on the EGLS framework. Delivered as code with complete governance declarations and audit artifacts. Clinical review at every release boundary.
Governed Artificial Intelligence for Medicine
Onde rigor clínico
rigor meets
machine learning.
5
Products in
portfolio
portfolio
3
US Provisional
Patents filed
Patents filed
3
Editor in Chief ·
Academic Editor roles
Academic Editor roles
70
Peer reviews
on ORCID
on ORCID
20+
Years of
clinical practice
clinical practice
I.
The principal
Julian Borges, MD.
Board certified endocrinologist and clinician scientist. Founder, CEO, and Chief Scientist of FxMEDUS, LLC. Editor in Chief at Aditum Publishing LLC. Academic Editor at Science Publishing Group. Former Academic Editor at PLOS Digital Health. Harvard Medical School GCSRT alumnus. JAMIA Open author.
At a glance
- Role
- Founder, CEO, Chief Scientist, FxMEDUS, LLC
- Training
- MD · MS Medical Genetics · GCSRT Harvard Medical School · MS Health Informatics, BU (2027)
- Certification
- Endocrinology and Metabolism (SBEM/CFM/AMB) · Medical Nutrition (ABRAN/CFM/AMB)
- Editorial
- Editor in Chief, Aditum Publishing LLC · Academic Editor, Science Publishing Group · Former Academic Editor, PLOS Digital Health
- Faculty
- Associate Professor of Medicine, Afya Medical Post Graduation Institute
- Societies
- Endocrine Society · AMA · AMIA · APSA · ASN · ACSM
- ORCID
- 0009-0001-9929-3135
FxMEDUS operates on a single premise: applied artificial intelligence in medicine is only as sound as the clinician who architects and governs it.
The firm holds three US provisional patents covering governed learning systems, ships two platforms in production, and maintains an active research program on clinical AI governance published across SSRN, Zenodo, ChemRxiv, and peer reviewed journals including JAMIA Open.
Engagements are selected on clinical and scientific fit rather than volume. Every deliverable passes through a documented physician review stage before release.
II.
The portfolio
Five products, three patents.
Each product is a governed instantiation of the firm's research program on adaptive clinical AI. Two are live in production, one in beta, two in earlier phases. IP filed under the Externally Governed Learning Systems (EGLS) framework.
Live
Phase 3
Drug Discovery · AI
DrugSynthAI
Governance first AI pipeline for de novo molecule design targeting genetically defined mitochondrial defects. Sixty seven autonomous agents across five tiers. Three validated discovery campaigns complete. US Patent 64/018,624 filed with three amendments.
drugsynth.ai
Live
Phase 3
Physician Operating System
FxMED OS
Intelligent Physician Operating System. 19 service layers covering patient registry, scheduling, encounter management, clinical documents, consent, and AI scribe under explicit BAA governance. Architectural core covered by EGLS Patent 63/975,551.
fxmed-os.vercel.app
Beta
Phase 2
Medical Licensure · AI Tutor
MedBoardPRO
Adaptive licensure preparation engine for USMLE and related boards. Three tier Analyst, Strategist, Judge architecture with EGLS governance. USPTO Trademark filed (Serial 99721498). US Patent 64/012,574 filed with 27 claim specification.
medboard.pro
In development
Phase 1
Trained AI Companion
FxMED Advisor
Clinician trained AI companion. Second production instantiation of externally governed learning systems. Embedded inside FxMED OS as an operator specific AI layer, distinct from generic documentation scribes.
Learn more
MVP
Phase 1
Medical Education SaaS
AFMI Platform
Physician first educational SaaS platform. Next.js and FastAPI with Clerk authentication, Supabase, and Stripe billing. Multi tenant architecture with tenant scoped access control.
Learn more
Operational
Infrastructure
Agent Orchestration
MedClaw
Local first multi agent orchestration runtime. Cost aware model routing with Ollama as the primary inference path and cloud fallback. Access control enforced at the orchestration gateway. Supports the broader FxMEDUS portfolio.
Learn more
The research program
Governing Adaptive
Clinical Artificial
Intelligence.
Structural failure modes, auditability, and infrastructure for decision safety. A unified framework for clinical AI governance across engineering, clinical, and policy layers.
Pillar 01
Formal theory
Externally Governed Learning Systems. A formal model treating governance as a viability constraint on adaptive computation. Foundation of US Provisional Patent 63/975,551.
Pillar 02
Empirical audit
Shortcut learning and misclassification evading standard validation. Peer reviewed in JAMIA Open, January 2026 (Oxford University Press). Harvard GCSRT capstone.
Pillar 03
Deployed infrastructure
Interoperable HL7 FHIR native governance infrastructure. Adaptive model selection via multi armed bandits. FxMED OS and DrugSynthAI as production instantiations.
Precision drug discovery platform
Governed AI for rare
genetic disease
drug discovery.
A general purpose de novo molecule design platform for patient populations the pharmaceutical industry has systematically deprioritized. Governed pipeline, validated architecture, and an open standard for responsible AI driven discovery.
Pillar 01
Multi agent architecture
Sixty seven autonomous agents across five tiers — pipeline, validators, orchestrators, intelligence, and precision medicine. One hundred sixty seven API endpoints. Five hundred thirty six passing tests including thirty seven security tests.
Pillar 02
Validated architecture
Build discipline under numbered rules R01 to R103. Three complete discovery campaigns executed. One hundred sixty three candidates ranked per campaign. Eighteen stage gate decisions, all passed. Zero kill switch activations.
Pillar 03
Open governance standard
AIDD-GOV v0.1 under Apache 2.0 — an open specification for AI drug discovery governance. Ten formal schemas, three conformance levels. US Provisional Patent 64/018,624 filed. Repository public on GitHub.
End to end proof of concept
MitoCoreX validated
the platform
end to end.
The first validated campaign run on DrugSynthAI. Eleven paper research program targeting five priority mitochondrial proteins through the full pipeline: architecture, druggability, variant modelling, ADMET, and in silico pharmacological profiling.
Pillar 01
Target architecture
Systems level map of mitochondrial pathway connectivity for precision drug design. Druggability assessment of structure function constraints and binding site characterization. Variant modelling to classify functional defects.
Pillar 02
De novo design
AI assisted de novo design of small molecule candidates against five priority mitochondrial proteins. Pan mitochondrial privileged scaffolds identified: adenine, pyridine, and a 7H purine bioisostere as multi target fragments.
Pillar 03
Pharmacological profiling
In silico ADMET, target engagement, selectivity, and stability analysis across the compound library. Published across Zenodo, ChemRxiv, and Research Square. Target journals include Briefings in Bioinformatics, Frontiers in Bioinformatics, and CSBJ.
III.
Selected publications
Recent scholarly output.
A selection from the 31 scholarly papers on SSRN and 27 DOIs across platforms. The complete record is maintained on the Publicações page and cross referenced to ORCID and Google Scholar.
2026
Peer reviewed
Auditing Shortcut Learning and Misclassification in AI Based Breast Cancer Genomic Subtyping
JAMIA Open Oxford University Press, Volume 9, Issue 2, April 2026
2026
Governance program
Externally Governed Learning Systems: Adaptive Computation Under Viability Constraints
SSRN Foundation paper of EGLS Patent 63/975,551
2026
Peer reviewed
Clinical Artificial Intelligence as a Sociotechnical System: Structural Failure Modes and Governance Requirements
Am J Biomed Sci & Res Volume 30, Issue 1
2026
Governance program
Adaptive FHIR Native AI Governance for Clinical Decision Support
Research Square Preprint · dataset on Zenodo
2026
DrugSynthAI program
MitoCoreX: A Validated AI Pipeline for Precision Drug Design Targeting Genetically Defined Mitochondrial Disease
Zenodo End to End Proof of Concept, v1.0
2026
Governance program
Auditable AI for Genomic Equity: A Compliance as a Service Model for Responsible Innovation
SSRN January 2026
2024
Endocrinology
The Inverse Association between Testosterone Replacement Therapy and Cardiovascular Disease Risk: A 25 Year Systematic Review
Int. J. Cardiovascular Medicine
Editorial leadership and peer review
Editor in Chief at Aditum Publishing LLC. Academic Editor at Science Publishing Group. Former Academic Editor at PLOS Digital Health.
Seventy verified peer reviews on ORCID, distributed across Public Library of Science (37), Oxford University Press (16, including JAMIA Open and EHJ Digital Health), Clarivate (14), Elsevier (2), and Springer Nature (1, Nature Reviews). Panel reviewer for AMIA 2026 and Endocrine Society 2026.
IV.
Speaking and conferences
Invited speaking.
Selected invitations and panel participations. Gold markers indicate upcoming engagements, burgundy markers past presentations.
-
2027 · April 13 to 17ACMG Annual Clinical Genetics Meeting — Clinical Genetics and AIAmerican College of Medical Genetics · Minneapolis
-
2026 · Decision 29 MayMEDevice Boston 2026 — SaMD Track: Governing AI in Drug DiscoverySpeaker proposal submitted · ten slide deck delivered
-
2026 · 7 to 9 MayAIMed 2026 Krakow — Encore Poster PresentationDrugSynthAI M6 ChemRxiv preprint · application 1400
-
2026 · 20 to 21 MarchSET III Boston University — Invited Speaker: Auditable AI for Genomic EquityBoston University
-
2025 · 20 to 21 NovemberICAIC 2025 Osaka — Invited Expert: Wave Equations in AI and HealthcareInternational Conference on Artificial Intelligence in Cybernetics · Osaka
V.
Engagement model
Three lines of work.
The firm accepts a small number of engagements each quarter. Selection is on clinical, scientific, or regulatory fit. Physicians and principal investigators receive direct founder contact.
Line 02
Clinical operations software
Patient facing applications and physician operating systems. FxMED OS available under selective partnership. HIPAA aligned deployment under an executed business associate agreement.
Line 03
Research infrastructure
Structured evidence synthesis, preprint triage, and literature surveillance pipelines for principal investigators and academic departments. Methods published and reproducible.
Many clinically consequential AI failures are system level phenomena arising from interactions among workflows, documentation practices, reimbursement incentives, and institutional accountability, rather than isolated algorithmic deficiencies.
VI.
Begin an engagement
For clinics, research, and press.
Inquiries reviewed within one business day. Identify the journal or institution in the subject line for editorial and scientific correspondence. Identify the outlet and deadline in the subject line for press.