I am a Machine Learning Engineer and Researcher specializing in Responsible AI and Low-Resource NLP for the Global South.

Currently, I work at TrustStamp, where I design secure identity systems and evaluate Large Language Models (LLMs) for hallucination and faithfulness in production. I am also a graduate of Carnegie Mellon University (MS ‘22), where I focused on speech recognition for East African languages.

My research goal: I am working on building safe, culturally aligned LLMs for critical domains in Africa. I am interested in how we can measure safety and helpfulness in lower resource high impact languages without compromising utility.

I am currently applying for PhD positions for Fall 2026.

Current Research Focus: Safety for African LLMs

As LLMs become the first line of defense for mental health support in Africa, we risk deploying models that are culturally misaligned or unsafe. I am currently designing benchmarks to measure safety, helpfulness, and calibration of LLMs in Swahili and Arabic, specifically for youth-typical scenarios.

News

• [Nov 2025] Applying for PhD programs in AI Safety & NLP.
• [Oct 2025] Paper accepted at ArabicNLP @ EMNLP 2025: “Enhancing Arabic Readability Prediction”.
• [May 2024] Published “Generating African Artistic Styles” at IST-Africa 2024.
• [Oct 2023] Joined TrustStamp as a Machine Learning Engineer.

Selected Experiences

Professional Experiences

• Trust Stamp: Built privacy-preserving, AI-based digital identity systems and strengthened model security through adversarial and watermarking techniques.
• RICOS Engineering: Designed renewable microgrids for refugee camps and reduced operational costs by 20% using data-driven optimization.
• Innovation Baylasan: Directed data science initiatives and developed forecasting tools for social-impact organizations.

Research

• Enhancing Arabic Readability Prediction with Hybrid BERT and Linguistic Features:
  • ArabicNLP Workshop @ EMNLP 2025 - Link
• Multilingual ASR for Kinyarwanda Language:
  • AfricaNLP workshop @ ICLR 2023 - Link
• Generating African Artistic Styles Using Textual Inversion:
  • IST-Africa 2024 - Link
• Enhancing Energy Trading Between Islanded Microgrids:
  • ICCCEEE 2021 - Link

Community Building

• Zindi.africa: Served as the Country Ambassador for Sudan.
• Deep Learning IndabaX: Organized and mentored teams for IndabaX Sudan and IndabaX Rwanda.
• IEEE Sudan: Led, organized, and advised on various initiatives to promote AI and machine learning.

Research Vision

Across all these projects, my goal is to design AI systems that are:

• Locally aligned with linguistic and cultural context

• Efficient and transparent deployable under limited compute

• Collaboratively governed co-developed with stakeholders, and communities