A learned validation model that approximates an expensive scientific or engineering simulator
Train a fast neural approximation on inputs and outputs from an expensive simulator, then use it as a validation device to screen many candidate configurations before committing to full simulation.
online B2B Deep Tech / Infrastructure
From Y Combinator — Jeff Dean: The 1% Rule for Building in AI at 44:32
Problem: A full-fidelity simulator may take a long time for each candidate, making large-scale screening impractical.
For: Scientific and engineering teams that need to evaluate large numbers of candidate designs or configurations but face slow, expensive simulations.
Examples
- Quantum chemistry: a neural approximation of a density functional theory simulator was described as being 300,000 times faster and nearly as accurate as the full-scale simulator.
Behind this: 11 build steps · 2 tools and how each is used · how to validate demand · 1 more real example · 5 things the video never answers.
Other takes on AI model development
- Retrain small open-source language models for long-horizon agentic workloads and sell efficient inference and on-device deployments
- AI simulation company that models human behavior and lets organizations test decisions against simulated populations
- An open foundation-model company focused on coding and long-horizon software tasks
- Enterprise context and training-data infrastructure for AI applications
- In-database semantic analytics using large database models
- A domain-specific AI product built around private data, specialized models, and purpose-built workflows