Machine Learning 49
- A Mine, Forty Clinics, Ten Million Downloads: What Open Weights Did for TB Screening
- Contamination Checks Before You Trust a Fine-Tune Score
- Fuse Ranks, Not Scores: Combining BM25 and Dense Retrieval
- 44,000 Seeds, One Open Map: What the Dryland Sorghum Pangenome Changes for Breeders
- Is Your +3-Point Eval Gain Real? Gating a Comparison with Paired Statistics
- Agreement Before Accuracy: Auditing the Labels You Are About to Train On
- The Model Was Not the Intervention: Reading Tanzania's Retina Screening Trial
- Your WHERE Clause Ate the Recall: Sizing the Candidate Pool Behind Filtered Vector Search
- Africa's Outbreak Radar Reached 44 of 47 States: Reading What the Expansion Buys
- Redact, Then Prove It: Per-Type Recall, Offset Damage and the Reversible Pseudonym
- The Bottleneck Is Adoption, Not Innovation: Reading Nigeria's AI Maternal Health Multiplier
- Your Load Test Lied About the Tail: Coordinated Omission and the Utilization Budget Behind p99
- Your Model's 0.9 Is Not a 90% Chance: Calibrating Risk Scores Before You Set the Threshold
- Sold to the Bank, Runs in the Bank: What Egypt's $13M Decisioning Round Actually Buys
- 300 Deaf Students, One Interpreter: What It Takes to Scale African Sign Language AI
- Prequalified: WHO Opens Its Procurement List to AI TB Screening Software
- Your Corpus Is Duplicating Itself — MinHash + LSH Dedup You Can Measure Tonight
- Four Times the Fields: The Delivery Stack Behind the Google–Gates Farmer AI Plan
- Nine Billion Variants, One Atlas: What DeepMind's AlphaGenome Atlas Makes Testable
- Thirty Times Smaller: Auditing Embedding Compression Before You Ship It
- The Mask Is the Contract: What Grammar-Constrained Decoding Actually Guarantees
- Nineteen Languages, One Download: Reading the TranslatePsy-AfriSLM Release Past the Headline
- Inconsistent by Construction: Reproducing and Catching a Repetition Collapse
- The Denominator Is the Metric: Auditing a RAG Retriever Before You Blame the Model
- Certified in Europe, Judged at the Health Post: What a Class IIb CE Mark Really Buys Primary Care
- Split Before You Believe: Why Offline Fraud Models Score Better Than They Perform
- The Lab Partner That Never Sleeps: How AI Now Designs Physics Experiments
- Fraud Rings Are a Graph Problem: Detecting Mule Networks in Transaction Graphs
- Fraud Models Rot Quietly: PSI, Feature Drift and Data-Quality Gates in Production
- Fraud ML in Mobile Money: Building Models That Catch the 70-Account Loophole
- AI at Industrial Scale: Dangote's $46B Bet and the Machine Learning Opportunity
- AI for Agriculture & Climate in Africa: Predicting the Unpredictable
- AI for African Fintech: Credit Scoring, Mobile Money & Fraud Detection
- AI in African Education: Personalised Learning for a Young Continent
- AI for African Healthcare: From Diagnostics to Drug Discovery
- Low-Resource NLP: Transfer Learning and Data Augmentation for African Languages
- The State of AI in Africa: Opportunities, Challenges, and the Road Ahead
- Building and Deploying AI in Constrained Environments: Low-Bandwidth and Edge Solutions
- AI for Agriculture in Kenya: Crop Detection, Weather Prediction, and Food Security
- African AI Communities: Masakhane, Lacuna Fund, and Local Dataset Initiatives
- AI for Mobile-First Markets: On-Device ML with TensorFlow Lite
- Swahili NLP: Building Language Models for African Languages
- Three New Series: Building AI Agents, African AI, and AI Infrastructure
- Knowledge Graphs Meet LLMs: RAG with Structured Knowledge
- Graph Neural Networks for Knowledge Graph Reasoning
- Knowledge Graph Embeddings: From TransE to RotatE
- Second-Order Optimization Methods: Beyond Gradient Descent
- Momentum and Adaptive Learning Rates: Accelerating Deep Learning Optimization
- Introduction to Gradient Descent: The Foundation of Machine Learning