Projects

Every published method ships with an implementation you can install and run. Papers that cannot be reproduced are opinions.

Research code
  • video-encoder-benchmark Python Elios Lab

    Reproducible hardware-aware benchmark harness for fifteen single and hybrid visual-backbone configurations across FP32/FP16/INT8, on NVIDIA H100 and Jetson Orin Nano. Companion to the JETCAS submission.

  • semideep Python pip install semideep

    Distance-based sample weighting for semi-supervised deep learning — the reference implementation of the IEEE ICMLT 2025 method. Drops into an existing training loop.

  • semicart Python pip install semicart

    Semi-CART: semi-supervised decision trees with distance-based weights, the reference implementation of the Springer IJMLC 2024 method.

Systems & experiments
  • cartgo Go

    CART in Go — fit, predict, save, load — written for low-latency serving paths where a Python round-trip is too expensive.

  • rust_decision_tree Rust

    Decision trees written from scratch in Rust, to see what tree learning actually costs when you control the memory layout.

  • NetExecGo Go

    Command-line tool for executing commands through a SOCKS proxy, for routing network work over a controlled path.

All repositories Write-ups on Medium