I build applied ML systems for low-resource language environments, neural interfaces, and autonomous systems — primarily for African contexts where real-world constraints define the design space, not benchmark leaderboards.

My work sits between language modeling, signal processing, and deployable infrastructure. The question I keep returning to: what breaks when you leave the lab?

African Language AI Platform  ·  Sema Labs, AWS Startup Program

Full-stack NLP platform for Central & East African languages with a specific focus on local dialect implementation — going below the national-language layer into the granular linguistic fabric of the region. Custom data pipelines for low-resource dialect corpora, cross-dialect translation infrastructure, and fine-tuned regional language models on scalable AWS cloud architecture.

Local Dialect AI Assistant  ·  Kenyan Language Systems

Voice and text AI system purpose-built for Kenyan dialects. Not an adaptation of an existing multilingual model — designed from the ground up for low-resource conditions with custom dialect training. Oriented around accessibility and cultural continuity.

Neural Signal Interface System

Experimental BCI featuring real-time EEG-style input classification. Evolved from a gesture-controlled interaction system into a cognitive-intent-driven control pipeline. Primary application target: assistive technology and human-computer fusion.

AI Phone Assistant  ·  LLaMA 3.1 + Local Dialect Voice

End-to-end conversational agent for voice calls. Powered by LLaMA 3.1 with local dialect voice synthesis — built for the phone call as Africa's primary communication channel. Not a chatbot adaptation; a purpose-designed call agent.

Autonomous Driving Simulation Engine  ·  AWS DeepRacer — Top 10% Global

RL pipeline for lane detection and real-time steering control. Applied reinforcement learning with computer vision for continuous decision-making under visual uncertainty.

Gesture-Controlled Virtual Keyboard

Real-time gesture recognition for touchless HCI. Open-source CV models applied to novel input — no physical peripherals.

Language & NLP       Low-resource modeling · dialect adaptation · multilingual transfer learning
Signal Processing    EEG classification · real-time neural input · biosignal pipelines
Computer Vision      Gesture recognition · lane detection · visual decision systems
Reinforcement Lrng   Continuous control · simulation-to-reality · policy optimization
ML Infrastructure    AWS · scalable data pipelines · production deployment
AI Safety            Technical safety fundamentals — BlueDot Impact (2025)

Tech writer at Techweez and Lighthouse — covering applied ML, African deep-tech, and the infrastructure behind the headlines. Focused on translating engineering complexity into precise, accessible writing.

Community Tech Lead — mentoring the next generation of African ML engineers, contributing to hackathons across the ecosystem, and working toward democratised access to ML tooling across the continent.

"Innovation is not just about technology, it's about solving real-world challenges" 🌍