AI Systems Engineering
AI systems that hold up in real use: reliable, observable and cost-aware, from retrieval pipelines to autonomous agents.
AI engineer and researcher building intelligent systems that work outside the lab.
I'm an AI engineer at Digital Umuganda in Kigali. I hold an M.S. in Engineering Artificial Intelligence from Carnegie Mellon University, where I was a graduate research assistant on business process optimization and a teaching assistant for the Agentic AI course.
Before graduate school I spent nearly three years at ITEC Ltd designing and running production software: payment systems and management information systems for businesses and institutions in Rwanda and across Africa. That background shapes how I approach real-world problems. Models matter, but so do the data pipelines, interfaces and failure modes around them.
AI systems that hold up in real use: reliable, observable and cost-aware, from retrieval pipelines to autonomous agents.
Sound architecture, testing and maintainable code that teams can keep building on for years.
Agentic AI and intelligent decision-making: how models plan, act and fail, and how to evaluate them rigorously.
Taking systems from prototype to production through evaluation, monitoring and performance at scale.
Joined Digital Umuganda as an AI Engineer, building and deploying AI-driven solutions.
Completed my M.S. in Engineering Artificial Intelligence at Carnegie Mellon University (GPA 3.87/4.0).
Began as Teaching Assistant for Agentic AI at CMU, mentoring students on multi-agent architectures and tool-using agents.
Started as a Graduate Research Assistant on business process optimization using AI techniques.
Building and deploying AI-driven solutions as part of the company's AI engineering initiatives.
Assisted with lectures, designed and graded assignments, and mentored students on agent-based systems, multi-agent architectures, tool-using agents and real-world AI applications.
Researched how business processes can be improved through the application of AI, including the development of new algorithms and the analysis of existing systems.
Built automation workflows and analytics dashboards for CMU-Africa's operations, applying AI to streamline business processes, improve reporting accuracy and support strategic decisions.
Designed, implemented and maintained enterprise applications, including the ITECPAY payment platform, improving system stability and backend efficiency.
Contributed to the design and implementation of software across client projects, with a focus on user experience and system performance.
Coded, debugged and documented core company projects, with hands-on exposure to enterprise-grade deployment.
Advanced machine learning, deep learning and intelligent systems engineering, with applications in natural language processing and computer vision. Research that bridges theoretical foundations with industry-relevant deployment.
Specialized in software engineering and distributed systems. Capstone: ophthalmology image translation, combining deep learning, computer vision and cloud deployment to support medical imaging and diagnosis.
Recent AI work first, then production systems, most of them proprietary or client codebases.
A team of agents (planner, reader, writer and critic) that turns a research question into academic writing with IEEE citations. Every quoted claim is checked against its source, references come from real bibliographic metadata rather than the model, and the critic's review triggers new searches and a revised draft. Outputs Markdown, IEEEtran LaTeX, BibTeX and a compiled PDF.
A benchmark for tool-using LLM agents: 34 sandboxed tasks across seven categories, from retrieval and error recovery to safety and long-horizon work, scored with deterministic checks for success, cost, steps and failure type. Compares native function calling against a ReAct baseline across models.
Thirteen Python packages for building LLM applications that keep working when things go wrong: document ingestion with OCR, hybrid retrieval, evaluation, query routing and agents, all behind one gateway to Gemini, OpenAI and local Ollama models.
Capstone research applying deep learning and computer vision to translate ophthalmology images into representations that aid diagnosis.
An integrated platform that automates inventory, invoicing and performance tracking, with AI-driven insights for day-to-day operations.
A machine learning system that forecasts price trends with time-series models over streaming data from financial APIs.
An interactive dashboard for visualizing sales, customer trends and operational KPIs.
A property management system for buildings sold or let unit by unit, with billing that reconciles to double-entry books.
A payment platform that processes transactions for businesses and individuals, built for reliability and scale.
A student information system that helps institutions manage student records, grading and attendance.
No projects match that search.
Recent essays from Medium.
Let's build something that matters.
I'm open to research collaborations, AI engineering roles and consulting on ML systems. Email is the fastest way to reach me.