Scale from a single developer and server to 100s of engineering teams and 10,000 nodes.

graph TD
    subgraph Deployment
        cloud("Cloud (AWS; Azure; Google; ...)")
        native("Native (Windows; Linux; UNIX; ...)")
        vm("VMs; Docker; …")
        wasm("WASM")
    end

    subgraph ML
        ml_python("Python compiler")
    end

graph TD
    subgraph Frontend
        android["Android (Kotlin [KMP])"]
        ios("iOS (Swift / KMP)")
        desktop("Desktop (KMP)")
        web("Web (KMP)")
        cli("CLI")
        sdk("SDK (C)")
    end

    subgraph Backend
        c("C")
        python("Python")
        rust("Rust")
        typescript("TypeScript")
    end

    Backend <-->|OpenAPI compilers| Frontend

From one [e.g., embedded] device to 10,000 servers:

• [old] 59+ Python repos with "off" prefix;
• [new] cross-platform [very] cross-platform package managers:
  • C89 base depending on each OS's crypto and network lib libacquire;
  • rvm/nvm style cross-platform package managers:
    • postgres-version-manager-go in Go;
    • "version-manager-rs" suffixed and verman-schema-rs Rust crates.
• [new] verMan.io for 1-click deploys: any {ML,database,server}; any {cloud,VM,docker,machine}; from/to any OS.

• Support for more cloud vendors;
• Uses normal Python packages deployable to PyPi, as opposed to Puppet/Chef/Ansible with their custom systems;
• [WiP] Deploy to any operating system (cross-platform: SunOS, Windows, Linux, macOS, OpenBSD);
• [WiP] Experiment with different versions of each package, including clustered variants.

From one cloud vendor to many:

• [old] See aforementioned Apache Libcloud and Fabric utilising Python repos;
• [new] C89 google-cloud-c library (soon: auto-generate entire library, and other vendors);
• [planned] autogenerate vendors other than Google Cloud.

• [C89] Can be called from most any programming language and runs in all environments;
• [planned] Build specific abstractions for multicloud, like: container-as-a-Service; ML-as-a-Service; Storage-as-a-Service; &etc.

From one machine-learning framework to many:

• [old] Python repo from my first PhD: ml-glaucoma;
• [new] 10+ Python repos with "ml-params" prefix:

• Keep up-to-date with latest innovations without porting to favourite framework;
• Experiment with every model on all major Python ML frameworks.

Compilers to automatically translate—within and—between:

• [intra-language] Automatically synchronise tests (& mocks), docs, types & interfaces;
• [exolanguage] Translate changes across language boundaries;
• Develop multi-language applications—e.g., Android, iOS, web, backend—as fast as single-language applications (compare with: Django or Ruby on Rails) and at a higher quality thanks to increased consistency, test coverage and doc coverage.