Onabanjo Michael
Lagos, Nigeria · Open to Work

Onabanjo
Michael.

I'm focused on how transport actually works, especially the informal systems that move most people across African cities. A big part of my interest is figuring out how to capture real-time data from these systems and use it to make transport safer, more predictable, and easier to navigate for the people who rely on it. At the core, I'm concerned with practical interventions, identifying key routes, integrating cleaner vehicle technologies, and building the planning frameworks that let informal and multimodal systems operate more safely and efficiently.

I serve on the Editorial Board of the IEEE Intelligent Transportation Systems Society Podcast, where I conducted the interview in Episode 82 on ITS activities across Africa and contributed to Episode 84 on cybersecurity in intelligent transportation. I am also a research assistant and member at TRATSEDI, collecting and analysing driver behaviour data to support transport research in West Africa, and participated in the Transformative Transport Service Conference 2025 in Loughborough. Before that, I spent a year simulating traffic on actual Lagos infrastructure. My microsimulation of the Third Mainland Bridge was calibrated to approximately 117,000 vehicles per day and showed a 20.72% reduction in emissions under a 10% autonomous vehicle penetration scenario. I hold a B.Eng. in Electrical and Electronics Engineering from Olabisi Onabanjo University, where I majored in Control and Instrumentation Engineering, and completed my undergraduate internship as a Control Engineer at Tulip Cocoa Processing Limited, a facility where over 90% of machines and devices are automated.

Work I've built

001 LIVE AI App · Deployed
NaijaTrip AI — Nigeria Travel Intelligence

A travel intelligence app covering all 36 Nigerian states and the FCT, combining an LLM with live traffic, weather, exchange rate, and advisory feeds to deliver route briefs, fare estimates, and safety guidance in real time.

React Groq LLM Node/Express TomTom API Vercel Prompt Engineering
002 MLOps · Production
Customer Satisfaction
Prediction to Production

Predicts customer satisfaction in e-commerce and segments outcomes into actionable risk tiers — green for automated approval, yellow for proactive outreach, and red for urgent intervention. The champion model, CatBoost, achieved a ROC-AUC of 0.72 and captured 71% of dissatisfied customers before they escalated.

Python Scikit-learn MLOps CI/CD Production
View Repository ↗
003 Deep RL · Lagos
Adaptive Traffic Signal
Control with DQN

A Deep Q-Network agent trained on Lagos intersection data in SUMO. Compared against fixed-time controllers, the RL agent achieved a 24% reduction in average travel time — replacing static signal timing with responsive, data-driven urban control.

Deep RL DQN SUMO TraCI Lagos
View Repository ↗
004 Research · Lagos
Third Mainland Bridge
— AV/HDV Simulation

Microsimulation of one of Africa's longest bridges, calibrated to ~117,000 vehicles/day. A 10% autonomous vehicle penetration scenario demonstrated a 20.72% reduction in traffic emissions — evidence-based intelligence for AV policy on critical Lagos infrastructure.

SUMO Python Autonomous Vehicles Emissions
View Repository ↗
005 AV · Simulation
AV Car-Following
& Lane-Changing

Models how autonomous vehicles make longitudinal and lateral decisions — following distance, speed adaptation, and lane-change triggering — and studies the effect on corridor-level traffic flow and safety.

Python AV Modelling Car Following Lane Changing
View Repository ↗
006 Logistics · EDA
Delivery Logistics Analysis
— Delhivery Network

Analyses Delhivery's logistics network to identify the key drivers behind delivery delays, surfacing patterns across trip types, geography, and time to give operations teams a clear basis for reducing late shipments.

Python Pandas Seaborn EDA Supply Chain
View Repository ↗
007 Logistics · AI App
Compliance QA Pipeline
— Logistics Video Intelligence

Logistics operations lose significant value to compliance failures that go undetected until it's too late. This system watches live video feeds from logistics facilities and automatically flags violations in real time — no human reviewer required. It reads the environment, cross-references the applicable regulations, and delivers a structured compliance report, turning what was a manual audit process into a continuous, automated one.

Azure AI Vision Azure OpenAI Azure AI Search RAG FastAPI Python OpenTelemetry
GitHub ↗

What I actually know

Transport simulation and modelling
SUMO Microsimulation Deep Reinforcement Learning TraCI Python API QGIS / GeoPandas AV Behaviour Modelling GTFS Data
Machine learning and AI engineering
TensorFlow / PyTorch Scikit-learn Stable-Baselines3 / DQN Hugging Face Transformers LLM APIs Computer Vision / CNN
MLOps and cloud deployment
AWS / Azure / Vercel DVC / MLflow / Kubeflow GitHub Actions / CI-CD Docker

Get in touch.

Open to research collaborations, entry-level ML and data roles, and conversations about transportation and AI in Africa.