Building infrastructure for systems
that can't afford to be wrong.
Voxell builds GPU-native infrastructure for real-time AI systems. We focus on the primitives that autonomous systems depend on: sorting, caching, rate limiting, and retrieval.
Our thesis: as AI agents scale to millions, probabilistic infrastructure creates compounding errors. Deterministic behavior isn't a feature, it's a requirement.
Founder & Lead Architect
GPU infrastructure engineer with deep experience in high-performance computing and AI systems.
Previously Lead Technical Architect for AI at a Berkshire Hathaway-backed enterprise. Creator of MASH Sort. 7 patents filed covering GPU sorting, caching, and AI infrastructure.
Building the infrastructure layer that will power the next generation of autonomous systems.
Probabilistic infrastructure creates problems that compound at scale. We build systems where the same input always produces the same output.
Not CPU algorithms ported to GPU. Everything we build is designed from scratch for massively parallel execution.
Microsecond latencies. Consistent performance under load. Built for systems where every nanosecond matters.
Not research prototypes. Infrastructure designed for 24/7 operation in the most demanding environments.
We're always interested in talking with teams building systems where performance and consistency matter.
Get in Touch