Tools and Projects

Talks are cheap, show me your code (Linus Torvalds).

Open-source toolkits I build and maintain. Issues and pull requests are welcome — feel free to ⭐ on GitHub if you find them useful.

Main Tools

Nkululeko

A high-level Python framework for predicting speaker characteristics from speech (emotion, age, gender, pathology, and more), with a one-config-file workflow for the full ML pipeline.

• 📖 Docs: nkululeko.readthedocs.io
• 💻 Source: github.com/felixbur/nkululeko

Speechain

Tools and utilities for speech processing and analysis — data prep, feature extraction, and training helpers reused across my speech experiments.

• 🌐 Site: bagustris.github.io/speechain
• 💻 Source: github.com/bagustris/speechain

Sherox

A CLI tool for speech AI inference. Run trained models on new audio files, realtime and offline, local-first, with support for batch processing and output formatting.

• Supported tasks: ASR, TTS, speaker verification, segmentation, language identification, and more.
• Source: github.com/bagustris/sherox

Coughkit

A Python package for cough sound analysis and classification with main focus on cough detection, segmentation, and counting.

• Source: github.com/bagustris/coughkit

Multibench

A benchmarking framework for evaluating and comparing multimodal models across various tasks and datasets, with a focus on human-AI interaction and real-world applications.

• 🌐 Site: human-ai-lab.github.io/multibench
• Source: github.com/human-ai-lab/multibench

PaperRAG

Local-first, CPU-friendly retrieval-augmented generation (RAG) over research papers — chat with your literature without sending it to a remote API.

• 🌐 Site: bagustris.github.io/paperrag
• 💻 Source: github.com/bagustris/paperrag

More

For other repositories, demos, and experiments, see my GitHub profile.