Robotics · HPC · AI · Privacy · Storage

Systems work across five interests, documented where it happens.

Every project page here is generated from the repository it describes — roadmap, pitfalls and lessons learned included, written in the repo and synced automatically. 6 projects currently tracked.

Browse the work GitHub

01Focus

0 projects Robotics Autonomy is a systems problem before it is an algorithms problem — power, thermals, and the bus usually decide before the model does. 1 project HPC Most clusters are not compute-bound. They are bound by the slowest thing between the accelerator and the data. 1 project AI Serving is where models meet reality: memory ceilings, cold starts, and the parts that do not fit on the card. 0 projects Privacy Inference you run yourself is the strongest privacy guarantee available, and it is getting cheaper every year. 1 project Storage Every tier is a bet about what you will need next. Caching is that bet made explicit.

02Featured

AI HPC Storage

Rustycage

A Rust LLM serving platform with block-level KV cache tiering across GPU, RAM, and disk

Active Rust today

ollama-kv-cache-tiering

Tiered KV-cache for Ollama: GPU VRAM → Host RAM → Disk, enabling near-infinite context length for LLM inference on consumer GPUs

Maintenance Cuda 6 months ago write-up pending

aiperf-lab-tests

AIPerf lab benchmark artifacts and test runs

Maintenance 4 months ago write-up pending

03Recently active

integrated-robotics

Isaac-Nexus: An Integrated Robotics Platform for Industrial Automation

Maintenance Mermaid 12 months ago write-up pending

kv-cache-tiering-rust

A high-performance KV cache tiering system implemented in Rust with intelligent data placement and compression

Maintenance 12 months ago write-up pending

All 6 projects →