Medicine × AI for Science
A clinician-led team of 19 clinicians, scientists and engineers across 8+ hospitals, universities and technology companies in China — building trustworthy AI for medicine and the life sciences.
Clinician-led, engineering-driven
AI4Viaevum was founded by Dr. Long Lai, an oncologist at Zhongshan People’s Hospital, to bring real clinical problems and modern AI to the same table. Our members practice oncology, hematology, radiology, urology, gastroenterology and general medicine — and build alongside engineers and research scientists from universities, AI labs and technology companies.
We compete to keep ourselves honest: international challenges in medical imaging, AI for Science and quantum computing give our work independent, measurable benchmarks. And we build for the long term — open-source code, reproducible pipelines and evidence-chained publication are defaults, not aspirations.
Clinical First
Every project starts from a real ward, a real cohort, a real unmet clinical need — not from a benchmark in search of a story.
Open Science
We publish methods, open-source our competition code and design for reproducibility from day one.
Full-stack Engineering
From data curation to Docker-submitted models and agent infrastructure, we ship working systems, not slide decks.
Competition record, 2025–2026
Independent benchmarks keep a young team honest. Selected results below.
CCF “Origin Cup” Quantum Computing Programming Challenge
Top award among 734 teams — lattice-based post-quantum cryptography and quantum algorithmic attacks, 2026.
2nd “Xingzhi Cup” National AI Innovation & Application Competition
National second prize in a field of 6,000+ teams, 2025.
ITU International AI Competition
International bronze award; ceremony attended in Switzerland, 2025.
1st “Shenzhi Cup” AI Innovation Competition
1,451 teams from 30+ countries; results released at WAIC 2026.
4th World Science Intelligence Competition
Semifinal Top-20 on four tracks — AI4S agents, biological structures, ancient scripts and nuclear fusion — among 17,977 participants from 32 countries.
BraTS Challenge — Brain Tumor Segmentation
Two teams (FLAIR-Viaevum & NeuroForge) with papers accepted on four tracks (T1/T2/T3/T5); #6 of 34 on the pediatric tumor validation leaderboard; code open-sourced.
GOAI 2026 — Global Open-source AI Challenge
Seven Viaevum squads across multi-agent infrastructure, AI+finance/education and AI4S frontier tracks; all initial submissions delivered, semifinal list pending.
What we are building
Four directions, one loop: from clinical questions to verifiable science.
Medical & Life-Science AI
ANIMA
FlagshipOrganoid Digital-Twin Operating System
An intent-driven digital twin for organoid research: state a question in natural language and the system orchestrates virtual knockouts, GRN simulation, scGPT perturbation prediction, virtual drug screening and causal inference — producing auditable R&D assets.
CellForge AI
Virtual-cell decision platform for drug-resistance research
Traceable virtual cells plus active learning compress massive intervention hypotheses in prostate-cancer resistance research into an interpretable, verifiable experiment queue.
Neutralizing Antibody Design
Generative AI vs. an unknown “Virus X”
Our entry in the 4th Bio-OS AI Open Competition (Guangzhou National Laboratory & ISCB-China): a literature-grounded, reproducible pipeline that designs neutralizing antibodies against an unknown viral target.
LifeMesh
HarmonyOS-based home health & safety coordination
Wearables, phones and home terminals form an auditable “self-confirm → family → institution” response chain for in-home risk events such as falls and abnormal inactivity.
AI for Science Infrastructure
VERA
Verifiable, executable, reproducible scientific publishing protocol
Submissions become a four-part bundle — structured paper, code repository, environment declaration and data manifest — with a built-in evidence chain. A six-layer pipeline covers AI-assisted intake, novelty checking, hybrid human-AI review, three-tier reproduction certification, layered reading with one-click re-runs, and a scientific knowledge graph with contradiction alerts.
Frontier Computing
Post-Quantum Cryptography
CCF “Origin Cup” challenge track
Designing lattice-based lightweight KEMs (192-bit+ quantum security, constant-time Cortex-M0 implementation) and developing quantum algorithmic attacks on lattice SVP/MLWE with real-device experiments and ML-KEM/ML-DSA security evaluation.
Agent Systems
LoopSeal
AI-native executable mind map
Natural-language intents, structured nodes, models, code, desktop automation and outcome evidence live in a single graph — closing the loop of intent → execution → verification → memory, with every side-effecting action requiring explicit authorization.
The people behind AI4Viaevum
Nineteen clinicians, scientists and engineers, organized in three groups — each with their own page.
Concept Team
Clinical concepts, study design and research direction.
Practice Team
Engineering, data and implementation.
Competition Members
Members who join the team for specific competitions.