AI Drug Discovery · Local · Privacy-First

We don't make drugs.
We make the starting points.

Novel drug-like molecules generated by a 2.27 billion parameter AI model, screened against validated pharmaceutical targets. Running entirely on consumer hardware. No cloud. No subscriptions.

2.27B
Parameters
79M+
Training Molecules
4.6M
Novelty Baseline
<30min
Full Pipeline
View Datasets → How it works ↓
The Platform

Built for real drug discovery research.

Every molecule screened against Lipinski Ro5, verified novel against 4.6M known compounds, and scored with real Ertl & Schuffenhauer SA scoring.

🧬

Real Novelty Verification

Every molecule checked against MOSES, ZINC250k, and ChEMBL. If it's in a known database, it's rejected.

⚗️

Real Docking Scores

AutoDock Vina against validated PDB crystal structures — EGFR (1M17), COVID-19 Mpro (6LU7), AChE (4EY7).

🔒

Privacy First

Everything runs locally. No API calls, no cloud compute, no data sent anywhere. Your research stays yours.

📊

Real SA Scoring

Ertl & Schuffenhauer synthetic accessibility scores. Not heuristics — the actual algorithm used in real drug discovery.

🎯

Drug-Like by Design

QED ≥ 0.50, SA ≤ 4.0, MW ≤ 500, full Lipinski Ro5 compliance baked into every generation run.

⚡

Consumer Hardware

2.27B parameter model running on a single RTX 5070. The most capable consumer-grade molecule generator available.

Datasets

Three tiers. One pipeline.

All datasets are delivered as clean CSV files. No account required. Guest checkout. Instant download.

Tier 1
Sample Pack
The 1,000 highest-scoring molecules from our generation pipeline, ranked by QED. A clean starting point for evaluation.
  • 1,000 novel drug-like molecules
  • Ranked by QED score
  • SMILES + QED + SA + MW + LogP
  • Lipinski Ro5 compliant
  • Novel vs 4.6M known compounds
$49 one-time
Download Sample Pack
Tier 2
Full Dataset
The complete generation output — 9,274 verified novel drug-like molecules with full property profiles.
  • 9,274 novel drug-like molecules
  • Full property profiles
  • SMILES + QED + SA + MW + LogP + ADMET
  • Top QED: 0.940
  • Lipinski compliance: 99.0%
$500 one-time
Download Full Dataset
Tier 3
Docking Dataset
144 elite compounds with validated AutoDock Vina binding scores against three major pharmaceutical targets.
  • 144 elite compounds
  • Real AutoDock Vina scores
  • EGFR · COVID-19 Mpro · AChE
  • Validated PDB crystal structures
  • Target-ready for further screening
$1,500 one-time
Download Docking Data
Secure checkout · Instant download · No account required · CSV format
The Pipeline

From generation to delivery in under 30 minutes.

A fully automated, reproducible pipeline running on local hardware. Every step verifiable.

Generate2.27B BPE Model
→
FilterLipinski Ro5
→
ScoreQED + SA
→
Verify4.6M Novelty Check
→
ADMETProperty Profile
→
DockAutoDock Vina
→
PackageCSV Delivery
🔐

Privacy-first from generation to checkout

The pipeline runs entirely on local hardware — no cloud compute, no API calls, no data transmitted anywhere. When you purchase, we use Stripe for payment processing and collect only what's necessary to complete your transaction. We don't sell your data. We don't track you beyond what's needed to deliver your download.