PhD Statement of Purpose for Nigerian Applicants

My doctoral research will develop new methods for resource-constrained machine learning inference, with specific focus on the deployment problem facing African industrial and agricultural sectors.

The research gap: Modern ML systems assume abundant compute, bandwidth, and electricity. African deployments lack all three. A maize disease classifier deployed in Kano cannot rely on a 5G uplink; it must run on a $40 Android phone with intermittent power. The literature on edge ML is rich, but the specific constraints of African deployment have received little attention.

My preparation: I hold a B.Eng. Electrical Engineering from the University of Lagos (First Class, CGPA 4.78/5.00, top of cohort) and an MSc Information Technology from Carnegie Mellon Africa (GPA 3.91/4.00). The formative experience was my 2-year postdoctoral position at the African Centre for Disease Control (Addis Ababa), where I worked on deploying ML-based disease surveillance systems across 8 African countries.

Published work: Two papers form the foundation. ‘Lossless Quantization for Edge ML on Intermittent Power’ (ICLR 2024 workshop, first author) and ‘Federated Learning Under Network Asymmetry in Sub-Saharan Africa’ (NeurIPS 2024 Datasets, co-author).

Why MIT, why this professor: Professor [Name]’s group at CSAIL is leading the theoretical work on resource-aware ML. Their 2024 paper on adaptive precision training provides exactly the framework I need.

5-year plan: Year 1: coursework + theoretical foundations. Year 2: first paper on rugged-edge inference. Year 3-4: system design + experimental validation. Year 5: thesis + job market.

After my PhD I plan to return to Africa as research faculty.