Skip to content
About

The white space incumbents leave open.

Generic AI drug-discovery platforms are heavy, therapeutically agnostic, and opaque. None of them are built for aging. GeroQubit occupies the gap they ignore: a rigorous multi-objective engine that is transparent, geroscience-native, and benchmarked against measured-lifespan data, with those results published in full, including the test that came back inconclusive.
Our principle

Measure, or don't ship.

A score earns its place only after it survives validation, and it always travels with its uncertainty.

Efficacy

A modest disclosed correlation, not a miracle.

Confidence

Calibrated on known actives, interval always shown.

Novelty

We say so when a chemotype is one the model has never seen.

Transparency is the differentiator. Our metrics are public, measured, and disclosed, including where they're weak.

What that cost us
Ideas we measured, then removed

Four of these five made our published numbers worse. We shipped them anyway.

A scoring axis that inflated every molecule

Near-constant, correlated ≈ 0 with lifespan. Dropped.

every score fell 6–9%
An accuracy figure built on our own data bug

Our standardizer handed leave-one-out a free perfect match.

ρ 0.182 → 0.158
A pharmacophore score that was nearly constant

It scored aspirin 0.958 against BCL-2. Rebuilt.

overall fell ≈ 7.5%
A credibility badge reading a bug

Novel molecules labelled "already known". Fixed.

0.900 → 0.662
More physics, which lost to less

3D shape + electrostatics, beaten by plain 2D.

0.572 vs 0.610

None of these were forced on us. Every one was found internally and could have been left alone. A number that only ever moves upward is not a measured number.

How it works

From target to synthesizable candidate.

01

Pick target + tissue

Built from real building blocks via named reactions. The route exists by construction.

02

Score on aging biology

12 hallmarks, measured-lifespan compounds, ADMET and binding plausibility.

03

Read the confidence

Every number ships with its interval and its disclosed accuracy.

04

Hand it to a chemist

Export with explicit routes and provenance. A starting point, not an oracle.

Deterministic and auditable. The same inputs give the same outputs on every run, with a registration hash per candidate. We cite the published methods we build on; our specific calibrations stay ours, and can be reviewed under NDA.

Who's behind it

Two chemists, building the longevity design tool they wanted.

Co-founder
Dinesh K
BSc Chemistry

Builds the platform. The de novo engine, scoring stack, and product.

Co-founder
Swetha H
MSc Chemistry

Drives the chemistry, medicinal-chemistry judgement and candidate evaluation.

Geroscience advisors, introductions welcome.

Trust, data & IP

What we build on, and what stays yours.

Public, citable data

DrugAge · ChEMBL · GTEx · López-Otín hallmarks. Lineage inspectable, not invented.

Nothing to set up

No install, no procurement, no training run to budget for. Send a target, get dossiers.

Deterministic & auditable

Same inputs, same outputs. Registration hash per candidate, versioned scoring.

Your chemistry stays yours

We never train on your inputs. Cloud, private VPC or on-prem; DPA available.