Kingsley Akenu, Data Scientist and MLOps Engineer

Live in Production

r-healthcare-readmission

Production healthcare ML pipeline with RAG-cited clinical decision support

LIVE IN PRODUCTION

End-to-end R pipeline that predicts 30-day hospital readmission risk from structured EHR data, explains each prediction with per-feature drivers, retrieves relevant clinical guidelines via a TF-IDF hybrid RAG system, and serves everything through a REST API with a full governance audit trail. Trained on 15,000 synthesized patients derived from real MIMIC-IV records. Every prediction is logged with a trace ID, model version, and hashed audit entry.

MIMIC-IV / synthpop → DuckDB / Backblaze B2 → XGBoost + RAG → Railway API → Shiny Dashboard

0.885

Recall (gate: 0.85)

0.566

AUC-ROC (honest)

71/71

Tests passing

8

Governance tables

GET /health Liveness probe

POST /predict Risk score + top drivers + trace ID

POST /explain Per-feature delta vs training median

POST /rag/summary RAG-cited discharge recommendation

R XGBoost tidymodels DuckDB RAG / TF-IDF Gemini API Plumber Shiny / Plotly Backblaze B2 Railway testthat GitHub Actions

Live Dashboard → API Health → Source Code →

crypto-price-pipeline

Production cryptocurrency forecasting system, built entirely in R

LIVE IN PRODUCTION

Ingests 5 years of daily Bitcoin and Ethereum prices from Yahoo Finance, engineers 15 technical indicators, fits ARIMA forecasting models with a temporal train test split, and serves predictions through a secured REST API to a live Shiny and Plotly dashboard. Retrains itself automatically every day at 2am UTC and flags model drift before it becomes a problem.

Yahoo Finance → R Pipeline / DuckDB → Railway API → Shiny Dashboard

0.0233

BTC RMSE

0.0358

ETH RMSE

35/35

Tests passing

Daily

Auto retrain, 2am UTC

GET /health No auth, liveness probe

GET /prices/history API key required

POST /predict/price API key plus rate limited

R DuckDB ARIMA Plumber Shiny / Plotly Docker Railway GitHub Actions

Live Dashboard → API Health → Source Code →

Experience

Data Scientist (Contract)

2025-03 → present

Kwurah, Ontario Canada, Remote

Leveraged analytics to align market trends with organizational needs, improving operational efficiency by 25%.
Developed Tableau dashboards for property market analysis, improving stakeholder decision making by 20%.
Used SQL and Python to process large datasets, optimizing strategic recommendations by 15%.

Freelance Data Analyst

2023-12 → present

Adezzz, Lagos Nigeria, Remote

Built an end to end workflow on Nigerian retail and e-commerce coupon usage data: Excel and MySQL for cleaning, Python for feature engineering, Power BI for delivery.
Built machine learning models (Logistic Regression, Decision Tree, Random Forest) reaching 60% accuracy, improving prediction reliability by 70%.
Delivered a predictive insights dashboard that improved coupon targeting efficiency by 30% and cut wasted marketing spend by 25%.

AI Trainer (Contractor)

2025-12 → 2026-03

Outlier (Scale AI), UK, Remote

Evaluated and ranked LLM outputs using RLHF principles to improve factual accuracy, reasoning quality, and safety.
Authored and edited prompts and responses, consistently meeting or exceeding project Quality Score benchmarks.