Senior Systems Developer

About Dweve Most of today’s AI is built around a simple assumption: bigger models, more compute, and increasingly complex infrastructure are the price of better intelligence. Dweve started by questioning that assumption. We are a European AI company building a different kind of AI stack from the systems layer upward. Our work combines neural learning with explicit constraints, deterministic execution, exact provenance and replayable evidence. Instead of treating explainability, efficiency and governance as features added after a model has been built, we design them into the way the system computes. That has grown into a much broader technology stack than a single model or AI product. Loom provides our specialist intelligence layer. Core contains the operations and execution substrate beneath it. Spindle handles knowledge, provenance and lineage. Nexus coordinates governed agents and organisations. Mesh distributes execution across infrastructure. Fabric is where people and developers actually work with all of it. Aura focuses on software engineering, while Charter handles the operational side around organisations and partners. Beneath that sits Kera, our systems technology for deterministic execution across different kinds of hardware. Alongside the commercial stack, we build open foundations that expose many of the lower-level ideas and contracts we work on. The common thread is simple: AI systems should be understandable, reproducible and practical to run. A developer should be able to know what happened, why it happened, what information was used and whether running it again will produce the same result. We are still a relatively small team, which makes the work unusually broad. People here are not maintaining one narrow piece of a mature platform. We are building the platform itself, from low-level execution and distributed compute to knowledge systems, agents and developer APIs. Dweve is based in the Netherlands, and builds primarily for European organisations and infrastructure. About the role A lot of what Dweve does eventually comes down to one question: how efficiently can we make the underlying computation run without giving up correctness? As Systems Engineer, you will work directly on that problem. You will join the team building Dweve Core, the execution layer underneath much of our stack. That means implementing low-level computation primitives, optimising them for different hardware targets, extending compiler paths and finding the places where a few instructions, a memory access pattern or a different representation can materially change performance. You will work closely with our Principal Systems Engineer, but this is not a role where you spend your time watching somebody else do the difficult work. You will own real parts of the system and be expected to make them better. Some days that means writing Rust. Others might involve SIMD intrinsics, GPU kernels, FPGA paths, assembly, compiler internals or staring at a profiler trace until you understand why something that should be fast is not. There is a lot to learn, and we are fine with that. We care more about strong fundamentals, curiosity and the instinct to measure before guessing than about whether you have already worked on every kind of hardware we support. What you'll do Implement new computation primitives across CPU, GPU and FPGA targets.Build and maintain low-level, performance-critical parts of Dweve Core.Benchmark implementations properly and understand where the time actually goes.Optimise memory access, vectorisation, instruction paths and hardware-specific execution.Work on compiler infrastructure targeting multiple execution backends.Maintain and extend our library of hardware-optimised algorithms.Profile real workloads and turn the results into concrete improvements.Work directly with the Principal Systems Engineer on architecture and implementation decisions.Own subsystems rather than only contributing isolated patches.Help make performance work reproducible, measurable and understandable to the rest of the engineering team. What we're looking for You should enjoy getting closer to the machine. You are comfortable thinking about what code turns into after the abstractions disappear. Cache behaviour matters to you. Memory layout matters. Branches, vector widths, data movement and instruction counts are not implementation trivia when they determine whether something runs well. You should be able to write solid systems code first and optimise it second. Fast but fragile is not particularly useful to us. Neither is beautifully abstract code that leaves half the machine unused. You do not need to arrive as an expert in CPUs, GPUs and FPGAs simultaneously. Very few people are. You do need to be willing to cross those boundaries and understand why an implementation that is excellent on one target can be completely wrong for another. You should have Strong Rust skills and experience with low-level systems programming.A solid understanding of computer architecture and how software interacts with hardware.Experience with at least one area such as SIMD, vectorisation, GPU kernels, assembly, compiler internals or hardware acceleration.The ability to benchmark and profile code rather than optimise from intuition alone.A strong bias toward correctness, reproducibility and measurable results.Excellent written and spoken English.Enough curiosity to keep digging when the profiler tells you something you did not expect. Dutch is useful, but not required. A degree is not required. Demonstrable ability matters considerably more to us. We are open to someone relatively early in a systems programming career if the fundamentals are unusually strong, as well as someone experienced who wants to work much deeper across hardware and compiler boundaries. What we offer A chance to work with technology that is still being invented rather than merely packaged.Direct access to the engineers and researchers building the systems you will work on.Significant ownership and influence over your area of Dweve.€5,000 per year for professional development.Competitive salary with an option package, primarily profit options.30 days of annual leave.Hybrid work in the Netherlands or Germany. Most importantly, this is a role for someone who finds it genuinely satisfying to turn “this should be faster” into a profiler trace, an explanation, a better implementation and a benchmark that proves it. If that sounds familiar, we should talk.