Domain-agnostic AI architectures — validated at AAAI, NeurIPS, Stanford — powering discovery across drug design, materials science, financial AI, and legal tech.
Enter password to view full site
Foundational architectures that transfer across disciplines — from molecular dynamics to regulatory compliance.
Integrates Bayesian Optimization with Generative AI and Causal Discovery to autonomously formulate, test, and refine scientific hypotheses. Reduces wet-lab iteration cycles by orders of magnitude.
Polyculture agent ensembles with built-in self-falsification and physics-informed surrogates. Every claim is verified against ground truth before propagation, eliminating hallucination cascades.
The same core methodology — validated independently in each domain with peer-reviewed results.
Accelerating drug discovery from target identification through lead optimization. ACHT-driven molecular generation with causal ADMET prediction.
Echo chamber detection, causal risk propagation, and verified portfolio construction. Separating signal from herding in markets.
Neuro-symbolic regulation models combining formal logic with LLM reasoning. Automated compliance verification with provenance tracking.
Physics-informed surrogates for crystal structure prediction, alloy optimization, and ceramic formulation with DFT-validated accuracy.
Selected papers from top-tier venues. Full list on the publications page.
ACHT Framework · Oral Presentation (top ~1% acceptance rate)
VAP Architecture · Formal verification for agentic systems
Materials Science Track · 100-1,000x acceleration over DFT
Stanford AI+Finance Workshop · Causal herding identification
AI & Law Track · Automated regulation verification
ELLIS Workshop on Reliable AI · Cross-domain methodology
Six platforms translating research into real-world impact — each validated against domain benchmarks.
End-to-end molecular generation with causal ADMET filtering and synthesis planning.
Multi-agent literature review, hypothesis generation, and experimental design.
Autonomous scientific discovery with formal verification and reproducibility guarantees.
Surrogate-accelerated ceramic formulation and firing profile optimization.
“Self-funded. Not-for-profit.
Radically open.”
Every architecture, dataset, and benchmark is published openly. We believe scientific acceleration requires institutional-grade rigour without institutional gatekeeping. No venture-backed pivots. No data moats. Just reproducible science.
From molecular discovery to climate science — explore the platforms built on ACHT and VAP.