Hover the data points, they scatter
Moving millions of passengers a day is one challenge.
Measuring it accurately is another.
For five years I built the analytics behind Lagos’ public transport, from databases to real-time dashboards.
Today, I’m researching computer vision for intelligent transport systems.
I work at the intersection of data science, machine learning and artificial intelligence, building models that see, read and reason about the real world, then turning what they learn into decisions people can act on.
About
That path started somewhere unlikely: a B.Tech in Transport Management Technology, where I learned how transport systems work long before I learned how to measure them. At LAMATA, my role quickly expanded beyond reporting. I modernized data operations by moving workflows from spreadsheets to a relational database, built a GPT-powered chatbot to enable non-technical staff to query data in plain language, and designed the cost model behind a fare review that cut First-and-Last-Mile operating costs by about a fifth.
Those experiences led to international collaborations, including the EU Horizon eBRT2030 programme, the EPIC UChicago Air Quality Fund, and remote machine learning engineering with teams in Zagreb and San Francisco. I’m now a Master’s research student at the Nagoya Institute of Technology, developing computer vision methods to track passengers in crowded, real-world transit environments.
Experience
Ozlab · Nagoya Institute of Technology · Japan
Researching "Unique Identifier based Segregation and Tracking across Video Frames" for automated bus passenger counting. Developed a large annotated dataset and conducted detailed analysis to improve tracking accuracy. Currently expanding the dataset for further research.
LAMATA · Lagos Metropolitan Area Transport Authority · Nigeria
Led the shift from spreadsheets to a relational database, built real-time dashboards and a GPT-powered analytics chatbot, and designed a cost model that cut First-and-Last-Mile operating cost by ~20% and lifted data-gathering efficiency by 80%.
Remotasks · Remote (San Francisco, USA)
Trained and evaluated model responses through iterative refinement, reviewed generated code for correctness, and hit 100% accuracy teaching a model to answer factual questions directly from CSV datasets.
Miksi · Remote (Zagreb, Croatia)
Built an SQL agent with a custom forecasting function answering natural-language questions over relational databases, plus a data-validation layer that infers datatypes and flags inconsistencies.
Enugu Electricity Distribution Company (EEDC) · Nigeria
Designed cost-effective dispatch route plans adopted team-wide and produced data-driven monthly operational reports.
Where it started Sales Analyst at Forever Living Products · Intern at Forte Oil Plc
Selected work
Automated passenger counters lose people in crowded buses, handing the same rider a new identity and inflating the count. My ongoing research investigates why trackers drop identities in dense, real-world transit footage, and how appearance, motion and association can work together to hold on to a person through occlusion, crowding and re-entry.
A GPT + LangChain interface over LAMATA operations data, letting non-technical staff query KPIs in plain language, no SQL required.
An XGBoost model ranking optimal electric Bus Rapid Transit corridors in Lagos from congestion, land-use, density and accessibility data.
A hybrid ML + deep-learning classifier for brand-reputation and customer-feedback monitoring with strong real-world accuracy.
A simulation framework forecasting Microsoft (MSFT) prices with probabilistic risk bands to support investment decisions.
A Power BI dashboard surfacing live ridership, revenue and service KPIs across LAMATA operations.
An AI SQL agent (Miksi) answering natural-language questions over relational databases with a custom forecast function.
Toolkit
Credentials
Let’s connect
I’m open to research collaborations, AI/ML roles and consulting. The fastest way to reach me is email.
esandamilola01@gmail.com