01 / Lab

Building research and software for intelligent systems.

Papers, notes, and the tools behind them.

02 / Focus

Three overlapping tracks: how things run, how they learn, and how we describe what we did.

  • a.

    Computing systems

    Inference paths, memory behavior, and runtime measurements on real hardware.

    • Inference
    • Compression
    • Profiling
  • b.

    Learning methods

    Models and representations where the structure is explicit and the evaluation is honest.

    • Sparse structure
    • Architecture
    • Evaluation
  • c.

    Practice

    Method notes, shared artifacts, and writing — described well enough to build on.

    • Notes
    • Code
    • Papers

03 / Software

What we ship

1 product

  • Internal · early access

    cortexSDR

    C++17 toolchain for compressing neural models into sparse .sdr archives and running sparse / on-demand inference from them. Supports ONNX, GGUF, TensorFlow, PyTorch, and HDF5, with BLAS- and SIMD-tuned kernels.

04 / Subscribe

Occasional notes when we publish.

No drip campaigns, no roundups, no marketing. Just an email when there's something new on the site.

Unsubscribe any time. We never share your address.