Our Technology Stack
At TensorFoundry, we believe in choosing the right tool for the right job. Our technology choices reflect our commitment to performance, security and developer happiness.
Programming Languages
Each language in our stack serves a specific purpose, chosen for its unique strengths and ecosystem.
Python
Core AI DevelopmentThe lingua franca of AI. Python powers our machine learning pipelines, data processing, and research prototypes. Its vast ecosystem of ML libraries and clean syntax makes it perfect for rapid development and experimentation.
Zig
System ProgrammingOur secret weapon for blazing-fast system components. Zig gives us C-level performance with modern safety features and zero hidden allocations. Perfect for our low-level tooling and CUDA kernels.
Golang
Backend ServicesGo powers our distributed services and APIs. Its goroutines make concurrent programming a breeze, while the simple syntax keeps our codebase maintainable. We write a lot of our CLI tooling and backend APIs in Go.
Rust
Performance CriticalWhen we need guaranteed memory safety without sacrificing performance, Rust is our go-to. It powers our most critical components where correctness and speed are non-negotiable.
C/C++
Hardware IntegrationThe foundation of high-performance computing. We leverage C++ for GPU programming, hardware acceleration and interfacing with existing ML frameworks. Essential for squeezing every bit of performance.
Svelte 5
Web DevelopmentOur choice for building beautiful, performant web interfaces. Svelte's compile-time and reactive programming model let us create fluid UIs that users love, without the runtime overhead.
Development Philosophy
Our engineering principles guide every line of code we write and every system we design.
Beautiful CLI Tools
We craft command-line interfaces that developers actually enjoy using. Clear output, intelligent defaults and helpful error messages make our tools a joy to work with.
High Performance
Every millisecond matters. We optimise relentlessly, from algorithm selection to memory layout, ensuring our systems run at the speed of thought.
Container-First
Born in the cloud, raised in containers. Our applications are designed from the ground up for containerised deployment, ensuring consistency from laptop to production.
Secure Defaults
Security isn't an afterthought-it's baked in. Safe defaults, principle of least privilege and defence in depth protect our users without getting in their way.
Maximum Customisability
Start simple, scale infinitely. Our tools work out of the box but offer deep customisation for power users who need to fine-tune every parameter.
Documentation First
Great code deserves great docs. We write documentation as we code, ensuring every feature is explained, every API is documented and every edge case is covered.
Security & Dependencies
Minimal Dependencies Approach
Every dependency is a potential security risk. We carefully evaluate each package, preferring to write critical functionality ourselves rather than inheriting unknown risks.
SBOM Generation
Every release includes a complete Software Bill of Materials, documenting every component and dependency.
Supply Chain Security
Signed commits, reproducible builds and dependency pinning protect against supply chain attacks.
Security-First Code
Input validation, secure defaults and defence in depth are standard practice, not afterthoughts.
Regular Audits
Automated security scanning and regular third-party audits keep our code secure and compliant.
Infrastructure & Deployment
Container Orchestration
Kubernetes-native from day one. Our applications scale horizontally, self-heal, and deploy with zero downtime. We leverage operators for complex stateful workloads.
- Horizontal pod autoscaling
- Rolling deployments
- Service mesh integration
- Custom operators for ML workloads
Edge Computing
AI at the edge of the network. We deploy models where your data lives, reducing latency and preserving privacy. From IoT devices to on-premise servers.
- Sub-10ms inference latency
- Offline-first architecture
- Adaptive model compression
- Edge-cloud synchronisation
Performance Optimisation
Every layer optimised for speed. From SIMD instructions to GPU kernels, from cache-friendly data structures to lock-free algorithms.
- Profile-guided optimisation
- Custom memory allocators
- Vectorised operations
- Adaptive batching strategies
CI/CD Pipeline
Continuous everything. Every commit triggers tests, every merge deploys to staging, every release is automated. Ship fast, ship safe.
- Automated testing pyramid
- Progressive deployments
- Automatic rollbacks
- Performance regression detection
Open Source Commitment
Building in the Open
Open source isn't just about code-it's about community. We believe in transparent development, collaborative innovation and giving back to the ecosystem that enables our work.
Olla
Our flagship open-source project. A local LLM runtime that makes AI accessible to everyone.
Learn moreContributing Back
We contribute patches, features and bug fixes to the projects we depend on.
Transparent Development
Our roadmap is public, our issues are open and our commits tell the story.
Ready to Build Something Amazing?
Join us in crafting the future of AI infrastructure. Whether you're deploying models at scale or building the next breakthrough application, we have the tools you need.