FigureAsia 35 Under 35 · Healthcare
Adil Kabylda
Age 29 · AI-enabled molecular simulation · Luxembourg
First and corresponding author of SO3LR, a model trained on four million molecular complexes and demonstrated at approximately 200,000 atoms on one GPU.
- Approximate age at 31 December 2025
- 29
- Field
- Healthcare
- Country or region
- Luxembourg
- FigureAsia U35 Assessment
- 86.2 / 100
Profile
Career and documented record
Drug discovery depends on simulations that are either fast enough to scale or accurate enough to trust; achieving both is difficult. Adil Kabylda's doctoral work introduced SO3LR, a quantum-informed machine-learning force field designed for realistic biomolecular systems. Published in 2025 with Kabylda as first and corresponding author, the model was trained on four million neutral and charged molecular complexes.
The reported system scaled to approximately 200,000 atoms on a single graphics processor and was applied to proteins, glycoproteins and lipid bilayers in explicit solvent. On a polyalanine benchmark, the paper reported an eightfold improvement in force error over AmberFF while retaining the scalability required for large systems. In 2026, Kabylda also contributed the QCell dataset, extending the open infrastructure around AI for molecular science.
This is enabling research, not a medicine. Its healthcare value depends on what future teams can discover or design with it, and performance will vary outside the chemical space represented in training data. Kabylda's leadership is nevertheless unmistakable: he is the lead and contact author, his doctoral thesis is built around the platform, and he has released data that lets others test the claims.
FigureAsia selection
Why Adil Kabylda is on the list
FigureAsia selected Kabylda because a Kazakh scientist has delivered a computational tool with credible global relevance and an auditable technical record. The scale, open data and authorship clear the bar for individual agency. The score does not pretend that better simulation has already improved a patient outcome; it recognizes infrastructure capable of changing the cost and reach of biomolecular research.
Verified work
The 2025–26 record
Principal milestone
Four million molecular complexes in the training data
Evidence record
Approximately 200,000 atoms simulated on one GPU
Scale or implementation
Eightfold force-error improvement on the reported polyalanine benchmark
Field context
The work in its field
Within ai-enabled molecular simulation, the relevant test is whether a result can survive scrutiny of maturity, attribution, validity and practical fit. That distinction matters: completed evidence is not projected benefit, and individual responsibility is not interchangeable with the wider team’s achievement.
FigureAsia U35 Assessment
Assessment breakdown
86.2out of 100
Substantive 2025–2026 contribution
20 / 20
The score reflects completed 2025–26 work in ai-enabled molecular simulation, assessed at the documented maturity of foundational computational research.
Verified impact
10.5 / 15
Impact credit is limited to the measured study, regulatory, implementation or operating record stated in the profile; unsupported patient benefit is excluded.
Originality and distinction
10 / 10
The work creates or materially advances a distinctive capability within ai-enabled molecular simulation rather than relying on title or institutional association.
Field and industry influence
8 / 10
The assessment recognises demonstrated effects on research, product development, care delivery or professional practice, with publicity alone carrying no weight.
Individual agency
10 / 10
Named authorship and the documented role of Computational Chemical Physicist establish individual responsibility while preserving credit for collaborators.
Durability and trajectory
4.5 / 5
The cited work forms part of a continuing programme, platform or research trajectory rather than a single uncompleted announcement.
Asian significance and global relevance
5 / 5
The Asian connection is material to the person's identity, operating base or populations served: Born in Pavlodar, Kazakhstan; completed his doctoral research in Luxembourg.
Clinical and scientific validity
5.6 / 7
Clinical and scientific validity is calibrated to foundational computational research, with the profile retaining the evidence boundary attached to the result.
Safety, quality and responsible governance
4.9 / 7
Safety and governance credit reflects accurate regulatory language, study limitations, data stewardship and the refusal to turn early evidence into clinical certainty.
Translation and care-pathway fit
4.2 / 6
The work is scored for its demonstrated fit with a laboratory, regulatory, clinical, operational or public-health pathway, not for projected future adoption.
Access, equity and resource stewardship
3.5 / 5
Access credit reflects documented reach, capacity, affordability or inclusion while distinguishing service volume from proven clinical outcome.