I am a Graduate AI Engineer at Ubundi working on the practical layer around AI agents: context, memory, internal tools, product workflows, evaluation loops, and the systems that make AI useful beyond the demo.
My work sits between product thinking and engineering execution. I like turning vague, messy workflows into structured tools, useful prototypes, and reliable operating systems for people and agents.
Current focus areas:
- Context-aware AI systems and retrieval workflows
- Agent infrastructure, memory, and tool orchestration
- Evals, observability, and trust in AI-assisted systems
- Internal tools that make teams faster without hiding the human judgment layer
- Physical AI data-quality infrastructure and trajectory QA
|
How AI systems remember, retrieve, and apply the right context without drowning users in noise. |
Turning one-off AI usage into repeatable workflows with tools, memory, verification, and clear human control. |
|
Making AI behavior inspectable enough to compare, debug, trust, and improve. |
Exploring how robotics trajectories can be checked, cleaned, explained, and trusted before model training. |
| Project | What it shows |
|---|---|
| Resonate | An AI-powered writing platform that models communication identity, rewrites outputs to match voice, and includes evaluation scoring and persona management. |
| Local Context Engine | A privacy-first local RAG app using offline document ingestion, local model execution, and hallucination evaluation. |
| Umbono AI Evaluation Dashboard | A dashboard for comparing AI model outputs against customizable evaluation criteria. |
| UbundiForge | AI-powered project scaffolding with team conventions baked in. |
The useful signal is in the pinned repositories and contribution graph below: current public work, practical AI tools, and the systems I am actively shaping.
I use AI heavily, but not as a shortcut around thinking.
The best systems still need human intent, taste, context, review, and accountability. I am interested in the engineering layer that makes that partnership real: tools, memory, evals, interfaces, feedback loops, and workflows people can actually trust.
More about my current learning lane
I am building toward stronger judgment in AI product engineering: how to scope useful agents, design context flows, evaluate outputs, protect trust boundaries, and turn research or vague workflow pain into working software. My current interests include agent harnesses, retrieval and memory systems, observability, workflow automation, and Physical AI data-quality tooling.
Currently building at Ubundi, learning in public where I can, and turning messy context into useful systems.