About Navatala

Navatala Systems builds two complementary things. First, we build open multi-backend GPU compute and AI libraries so developers can run scientific, data, graph, ML, and CFD-style workloads across more than one GPU ecosystem. Second, we apply that compute foundation to a marine digital twin spanning ship design, inspection, operations, and compliance.

Our Approach

We use formal specifications and contract-driven generation to keep behaviour consistent across a large surface area. The public output is practical software: GPU runtimes and kernels, Python bindings, desktop/mobile applications, and domain workflows. The generator stays internal; the released artefacts are what users build on.

What We Build

Open-source GPU + AI kernels. navatala_gpu is a multi-backend GPU runtime and kernel corpus spanning AI/ML primitives, dataframe operations, graph analytics, vector search, sparse linear algebra, reductions, scans, and CFD building blocks. Apache-2.0, developer preview, on GitHub at github.com/navatala-systems/navatala_gpu.

Marine digital twin. A single model of the vessel across its lifecycle. Navatala Ship Design is the as-designed model (hull, structure, hydrostatics). Talika VIR is the as-inspected condition (walkthroughs, deficiencies, evidence). Talika FMS is the as-operated and as-reported state (voyages, fuel, EU FuelEU, IMO DCS).

Technology

Our technology stack is built on open standards and verified engineering practices:

• Formal specifications with machine-checked contracts
• Multi-platform application code generation (C++, Kotlin, Swift, Dart, TypeScript, C#)
• Multi-backend GPU code generation (CUDA, HIP/ROCm, Vulkan compute + SPIR-V, OpenCL, Metal)
• Offline-first architecture with delta synchronization
• PostgreSQL with RBAC and audit infrastructure
• Apache-2.0 open-source release of the GPU runtime and kernel corpus