cusp.ai provides a cloud‑based AI platform that generates and evaluates candidate materials using generative deep‑learning models combined with physics‑informed property predictors. The system offers high‑throughput virtual screening, active‑learning loops that incorporate experimental data, and API/SDK access for seamless integration into R&D and manufacturing workflows, enabling faster identification of high‑performance compounds.
Funding
$100M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
NTHFounders
Product
Problem
Materials research traditionally relies on trial‑and‑error synthesis and high‑cost computational simulations, resulting in development cycles that span years to decades and limiting the pace of industrial innovation.
Solution
cusp.ai offers a cloud‑based AI platform that automates the generation and evaluation of candidate compounds, reducing the time required to identify high‑performance materials from months to weeks. The system combines generative deep‑learning models with physics‑informed property predictors to propose chemically viable structures and estimate key performance metrics. Users can iteratively refine designs through an active‑learning loop that incorporates experimental feedback, accelerating convergence toward target specifications. Results are delivered via a secure dashboard and API, enabling seamless integration with existing R&D workflows and downstream manufacturing pipelines.
Target Audience
Primary customers are R&D teams in chemicals, energy storage, aerospace, and electronics firms, as well as academic laboratories focused on accelerated materials discovery.
Features
- Generative neural networks that create novel molecular and crystalline structures conditioned on user‑defined property targets
- Multi‑task property prediction models (e.g., conductivity, stability, hardness) trained on curated experimental and simulation datasets
- High‑throughput virtual screening pipeline that ranks millions of candidates using surrogate models before optional DFT refinement
- Active‑learning framework that updates model weights with new experimental data to continuously improve prediction accuracy
- RESTful API and SDKs for programmatic access, allowing integration with laboratory automation and PLM systems
- Cloud‑native data lake with versioned datasets, provenance tracking, and role‑based access controls for collaborative projects
- Automated report generation with visualizations of structure‑property relationships and suggested synthesis routes