# Makini AI

**Open-source AI for Africa.** Operated by Algedi Intelligence Labs, Nairobi, Kenya.

AI is already deployed across Africa, but most of it was tested, tuned, and safety-checked in English, not in the languages people actually use. Makini AI builds open evaluation tools, datasets, and language resources so AI works accurately, safely, and fairly for African communities, in their own languages.

**Everything we build is free and open-source. No proprietary lock-in.**

---

### What we build

**Iroh** is a legal AI assistant grounded in the complete Laws of Kenya (546 Acts plus thousands of judgments). It gives accurate, source-cited answers where general models hallucinate. [Try it](https://huggingface.co/spaces/MakiniAI/iroh)

**Kenya Law Corpus** is an open corpus of the complete Laws of Kenya, structured into 63,000+ searchable chunks. It is the substrate for legal-AI research. [Dataset](https://huggingface.co/datasets/MakiniAI/kenya-law-corpus)

**MAKINI Bias Detection** detects stereotype, counter-stereotype, and derogation content in Swahili, French, and isiZulu; Hausa currently supports neutral-vs-biased detection only, with full coverage in progress. Built on AfricaBias — 60,234 annotated sentences across all four languages, the largest open African-language bias dataset. [Model](https://huggingface.co/MakiniAI/makini-bias-v2) and [Dataset](https://huggingface.co/datasets/MakiniAI/africabias-sw-fr)

**MAKINI Hate Speech** is a hate-speech detection model for Swahili and French, documenting where models trained on European data fail on African content. [Model](https://huggingface.co/MakiniAI/makini-hate-v1)

---

### Our commitment

We release our work as a public good, accessible to researchers, developers, governments, and communities across Africa.

[makini.tech](https://www.makini.tech) and [LinkedIn](https://linkedin.com/company/makinitechnologies)