Between the model and the system.
Machine learning is more than the model itself. It is also the data, the experiments, the evaluation, the infrastructure, and the decisions that determine whether a method works outside a notebook. I'm interested in the engineering that connects these pieces.
Useful machine learning begins with understanding the data: how it was collected, what it represents, what is missing, and how its limitations affect the questions we can ask.
Experiments turn ideas into testable questions. I am interested in clear hypotheses, controlled comparisons, reproducible workflows, and understanding why a method succeeds or fails.
Models are the computational core, but implementing them is only part of the work. Their behavior depends on the data, training process, assumptions, and environment in which they operate.
Evaluation asks whether a system actually works for the task and setting it was designed for. That means appropriate metrics, honest comparisons, careful validation, and transparent limitations.
Systems engineering connects individual components into something dependable. Interfaces, data flows, resource constraints, reproducibility, and failure handling all matter.
The final question is how technical capabilities become useful in practice. I am interested in the gap between promising research results and systems that address real problems.
Questions I keep coming back to.
These are the areas I want to investigate more deeply through implementation, experimentation, and research.
- Machine learning research engineering
- AI systems and ML infrastructure
- African speech and language technology
- Automatic speech recognition and text-to-speech
- Multilingual and low-resource NLP
- Dataset and language-resource development
- Model evaluation and reproducibility
- Applied machine learning and geospatial research