Adejare Adelugba
ML Engineer & Data Scientist
I build ML systems that work on real, messy data — not clean benchmarks. Most recently: 87.1% ROC-AUC on multimodal cancer detection, and electoral data infrastructure spanning all 36 Nigerian states.
By the Numbers
Measured outcomes from real work
Selected Work
Case Studies
The two projects that best represent what I build and how I think.
Multimodal Breast Cancer Detection
PyTorch • MedCLIP • TabNet • Early Fusion
Designed a fusion architecture combining ultrasound image embeddings (MedCLIP, 512-dim) with clinical tabular metadata (TabNet) to overcome the limitations of single-modality screening. Early fusion outperformed late fusion and both unimodal baselines.
class CancerFusionModel(nn.Module):
def forward(self, img, tab):
v = self.medclip(img) # 512-dim
t = self.tabnet(tab) # 64-dim
return self.head(
torch.cat([v, t], 1)
)
Nigeria Electoral Data Infrastructure
Python • Selenium • PostgreSQL • Click CLI
Built a multi-layer Selenium scraper navigating dynamically rendered state portals across all 36 Nigerian states. Standard scrapers fail at depth 3+ due to session token handling — solved with explicit wait chains, exponential backoff, and a resumable CLI.
StateWorker(state_id)
├── navigate_portal(depth=3)
├── retry_handler(backoff=exp)
├── extract_polling_units()
└── validate_and_clean()
│
bulk_insert → PostgreSQL
Skills & Competencies
What I build with
Mapped by real-world depth — tools I've actually shipped projects with.
Credentials
Certifications
Verified completions from Anthropic, AWS, and partner institutions.
Available for ML engineering and data science roles
Remote · Abuja, Nigeria · Open to relocation