Edge-AI for
under-resourced languages.

SautiLM develops offline NLP and Edge-AI architectures to restore mother-tongue education in the conflict zones of Eastern DR Congo.

Support our research

Context & Rationale

01. Context

Linguistic Exclusion

The digital divide systematically ignores vernacular languages, isolating entire cultures from technological progress.

02. Urgency

Crisis in Eastern DRC

In the conflict zones of Eastern DRC, the absence of local languages such as Bbadha from EdTech hinders the cognitive development of displaced children.

03. Vision

Edge-AI Resilience

We engineer autonomous models on solar-powered terminals, ensuring access to knowledge without Cloud connectivity.

Methodology & Engineering

Data Engineering

Multimodal Corpus

Digitization of bilingual dictionaries (starting with Bbadha-Swahili) and acoustic data collection via an ethical participatory approach.

Transfer Learning

NLP Modeling

Fine-tuning state-of-the-art self-supervised architectures (wav2vec 2.0, mBART-50) for extremely low-resource languages.

Hardware Optimization

Edge-AI Inference

Post-training model quantization (INT8) for low-latency offline execution on mobile devices.

Research Team

SautiLM is a technical research initiative led by Georges Byona & Abraham Imani, supported by LosingTek, a lab dedicated to solving humanitarian crises through technological innovation.