SPRINTER is an AI safety and research institute based in Nairobi. We study alignment, interpretability, and robustness — building the foundations for AI systems humanity can trust.
Researching methods to ensure AI systems reliably pursue intended goals — developing formal frameworks for value specification and goal stability under distributional shift.
Building tools that let humans understand what neural networks have learned — mechanistic analysis, circuit-level probing, and automated feature attribution for large models.
Strengthening AI systems against distribution shift, prompt injection, and adversarial inputs. Testing and red-teaming frontier models for edge-case failure modes.
Contributing to the policy discourse on responsible AI deployment in the African context — working with institutions to develop meaningful oversight frameworks.
Designing scalable oversight mechanisms and monitoring systems that keep humans meaningfully in the loop as AI capabilities grow.
Translating theoretical safety research into practical tools — shipping production-ready AI systems with safety built in from the ground up, for real clients today.
We build AI systems that reflect human values — not by hard-coding rules, but by studying how values can be robustly learned, specified, and preserved across a model's lifecycle.
Every system we build includes interpretability as a first-class requirement. We want to understand what our models are doing — and we want you to understand it too.
Oversight must scale with capability. We design monitoring and control mechanisms that remain meaningful as systems grow more capable — not just rubber-stamp checkpoints.
We stress-test systems before deployment — red-teaming, adversarial probing, and distributional shift analysis — so failures surface in the lab, not in the world.
AI safety research must reflect diverse contexts. We bring an African perspective to global alignment questions — ensuring future systems work for humanity broadly.
A 24-month research initiative to build the first comprehensive, open-source AI ecosystem fine-tuned for Swahili and East African languages — serving over 200 million speakers across six high-impact sectors.
Over 200 million people across East Africa speak Swahili, yet tools like ChatGPT, Claude, and Gemini perform significantly worse — sometimes catastrophically — when applied to Swahili and other African languages. Healthcare workers, farmers, students, and legal aid seekers are excluded from the AI revolution because of this linguistic gap.
AI-assisted triage and health information for rural East African communities and community health workers.
Swahili-language advisory tools for smallholder farmers improving yields and food security.
AI tutoring in Swahili for K-12 students learning in their native language.
AI tools supporting financial literacy and inclusion for unbanked populations via Web & USSD.
Accessible legal information and guidance for low-income Swahili speakers.
Cover letter, CV, and job application AI tailored to East African job markets.
Whether you're a researcher, organisation, or funder — we want to hear from you. Based in Nairobi, working globally.
airesearch@sprinter.co.ke · +254 704 445 453