AppCubic · Applied AI studio · Building since 2019
Agent workflows, research tooling, and the products that grow out of them: built, tested, and carried to production by a small studio that ships.
Focus areas
The work sits where prototypes have to become dependable: agent workflows, technical products, scientific tooling, and the operations around them.
Agentic systems
Portable skills for coding agents, document and web tooling, multi-agent orchestration, and local models for Apple Silicon. Built in public, shaped by daily use inside the studio.
Explore App AutomatonResearch to product
Scientific and analytical workflows turned into systems people can actually run, built for correctness and repeatability. One example is mlx-atomistic, an Apple-Silicon molecular dynamics and density functional theory runtime that runs on the Mac GPU.
Research systemsVenture & advisory
Early product strategy, technical diligence, grant support, and incubation for AI-first products.
About the studioOpen source
MIT-licensed tooling, public on GitHub and in daily use.
2019
Building since
Operating principles
AppCubic is a small AI engineering studio. We build, test, and carry agentic systems toward production with teams who know their domain deeply.
Agents earn trust through evaluation, careful retrieval, and explicit tool boundaries. We engineer for behavior you can predict.
We build end-to-end workflows shaped by real operating constraints, not isolated prompt experiments that break the first time conditions change.
Technical depth speeds up learning. Good data infrastructure, scientific rigor, and product judgment reinforce one another.