Build what comes next.
Most foundational software was designed for hardware that no longer exists. We're rebuilding it from first principles, engineered for how modern silicon actually works.
The Work
Problems worth solving.
At the lowest level.
We work on algorithmic infrastructure, the kind of software that sits beneath everything else. Compression codecs, data encoding, high-performance compute kernels across CPUs, GPUs, and the browser. The problems are hard, the constraints are real, and the impact is measured in billions of bytes.
Algorithm Design
Develop novel prediction models and transforms for scientific raster data. Prove correctness. Measure everything.
SIMD & GPU Engineering
Write code that maps directly to vector instruction sets and GPU compute shaders. CPU, GPU, and WebAssembly. Every cycle counts.
Systems Programming
Cache-aligned data structures, zero-copy pipelines, bit-level encoding. Performance is a design requirement, not an afterthought.
Cross-Platform Portability
Guarantee bit-exact output across CPUs, GPUs, and browsers. Native binaries to WebAssembly, server to edge.
Benchmarking & Validation
Design test suites across 14+ data types. Validate against real-world datasets. Prove every claim with numbers.
Developer Experience
Build SDKs, APIs, and integrations that make complex compression feel simple. Domain profiles, format drivers, documentation.
How We Work
Engineering principles,
not corporate values posters.
Measure, Don't Guess
Every design decision is backed by benchmarks. We profile before optimizing, test against real data, and let the numbers settle debates.
Depth Over Breadth
We'd rather understand one problem completely than skim across ten. If you enjoy spending a week on a single function to get it right, you'll fit in here.
Correctness Is Non-Negotiable
We build both lossy and lossless compression. In lossless mode, every decoded value must match the original exactly. In lossy mode, error bounds are guaranteed. There is no "close enough" without a contract.
Small Team, High Trust
No layers of management. No approval chains. You own your work from design through benchmarking. Ship when it's right.
Who We're Looking For
You see legacy infrastructure
as an opportunity.
We're looking for people who think in bits, care about cache lines, and get excited when a new approach shaves microseconds off a hot loop.
Formal credentials matter less than demonstrated ability. If you've contributed to open-source codecs, written SIMD intrinsics or GPU compute shaders, built something in WebAssembly, or spent time understanding how processors actually execute instructions, we want to talk.
Relevant experience might include:
- Systems-level C, C++, or comparable systems languages with a focus on performance
- SIMD programming, GPU compute (CUDA, Metal, WebGPU), or WebAssembly
- Data compression, signal processing, or information theory
- Geospatial formats, medical imaging standards, or scientific data pipelines
- Open-source contributions to compression libraries, codec implementations, or performance tooling
Interested in
working with us?
We don't have a list of open positions with bullet points. Tell us what you're good at and what you want to work on. If there's a fit, we'll figure it out.