Engineering & architecture

I design data platforms and build analytics engineering workflows on Google Cloud (GCP), from ingestion and BigQuery modeling to orchestration and operational applications.

Experience cases

Gozem Data Hub: operational apps & analytics platform

Data Director · GOZEM

Unified analytics, AI, and operational tools into a single platform serving 700+ stakeholders across five countries.
Challenge

As Gozem evolved from a startup into a scale-up, the demand for operational applications, automation workflows, analytics products, and AI-powered tools grew rapidly across the organization.

Product and engineering teams remained primarily focused on revenue-generating marketplace initiatives, limiting their capacity to support the growing number of operational use cases. Many requests to the Data team recurred, including ad hoc analyses, data extractions, and operational workflows that required intervention from analysts and engineers.

The Data team also lacked a shared environment for its growing portfolio of tools, applications, analytics assets, and AI services. It had two goals.

  1. Give operational teams direct access to data products, analytics, and tools to reduce recurring requests to the Data team.
  2. Establish a common platform for the Data team to quickly design, deploy, and govern applications, automation, and AI solutions at scale.
Approach

Architected and led the development of Gozem Data Hub, a centralized data and AI workspace for delivering self-service products across the organization.

The platform brought operational applications, analytics tools, AI solutions, automation workflows, and data exploration into one environment with shared standards for authentication, authorization, deployment, and governance.

The Data team used reusable architecture and AI-assisted engineering workflows to deliver production applications for operational teams. These applications moved a large share of analytical and operational requests from manual tickets to self-service tools that business users could operate directly.

Business teams could independently access insights, applications, and AI tools, and the Data team had more time for higher-value initiatives.

Tools

Enterprise metrics governance & AI-ready semantic layer

Data Director · GOZEM

Established a single source of truth for business metrics across analytics, reporting, and AI-powered workflows.
Challenge

As Gozem expanded its analytics footprint across multiple business domains and countries, the risk of inconsistent metric interpretation increased. KPI definitions, transformation logic, aggregation rules, dimensions, and query implementations were often embedded across dashboards, SQL queries, reports, and undocumented team knowledge.

At the same time, emerging AI-powered analytics workflows created new requirements. Business users needed trusted natural-language access to data, while AI agents required a reliable understanding of business concepts, metrics, and relationships to generate accurate insights.

The organization needed a foundation capable of serving both humans and AI systems with a consistent interpretation of business performance.

Approach

Designed and implemented the foundation of an enterprise metrics governance framework covering KPI definitions, source systems, transformation dependencies, aggregation behavior, available dimensions, lineage, and canonical calculation methods.

Created a centralized metrics catalog documenting how each KPI is derived, whether from raw operational data or transformed analytical models, including supporting business context, relationships between metrics, valid dimensions, joins, and calculation formulas.

Standardized KPI computation through reusable BigQuery and Dataform assets powering reporting and analytical applications, ensuring consistent metric definitions across dashboards, analyses, and operational workflows.

Introduced governed AI access to warehouse data through an MCP-enabled BigQuery integration in Claude. Selected senior users could query enterprise data using natural language within their existing Google Cloud IAM permissions.

This work also laid the foundations for an AI-ready semantic layer capable of supporting future copilots, autonomous agents, self-service analytics experiences, and automated decision-support systems.

Tools

AI-powered invoice processing for financing operations

Data Director · GOZEM

Automated monthly invoice reconciliation, with estimated savings of 40+ hours per country Financing Manager each month.
Challenge

Gozem’s vehicle financing business enabled drivers to acquire motorcycles and vehicles through installment-based financing. Across multiple countries, manufacturer dealerships and service partners performed maintenance and operational services and submitted monthly invoices covering hundreds or thousands of financed vehicles.

Before approving supplier payments, Financing Managers manually reviewed invoices, identified vehicle-level charges, prepared payment requests, and reconciled maintenance expenses against collections already received from drivers. They worked across invoice formats, built reconciliation tables, and cross-checked operational and payment records to verify that driver collections covered the charges before payment.

The process took Financing Managers approximately one week each month in every operating country.

Approach

Identified maintenance reconciliation as a high-impact automation opportunity and validated the extraction approach through rapid prototyping in Vertex AI Studio.

Designed and implemented an AI-powered document intelligence workflow capable of extracting vehicle-level maintenance records from scanned supplier invoices, then automatically matching them against financing and collections data already maintained by the data team.

Built an agentic workflow incorporating schema-driven extraction, validation rules, anomaly detection, confidence scoring, exception handling, and human-review mechanisms. Integrated the solution into existing payment preparation and reconciliation workflows through Gozem Data Hub, allowing Financing Managers to focus on reviewing and validating charges rather than manually transcribing invoice data and constructing reconciliation tables.

The solution transformed a process that previously required approximately one week of effort each month per country Financing Manager into a largely automated workflow, saving an estimated 40+ hours per month per country, accelerating payment-request preparation, strengthening maintenance-cost recovery controls, and improving the scalability of vehicle financing operations across multiple markets.

Tools

Operational command center for public events

Data Director · GOZEM

Supported near-real-time marketplace operations during Vodun Days. On the peak day, ride requests were 48% above the average for the other event days, and the rider match rate was 88%.
Challenge
As the official transportation partner for a public festival (Vodun Days), Gozem faced a high-visibility operational challenge in a city where it did not normally operate. Some drivers were mobilized from neighboring regions to support expected demand. Operations lacked historical references, and the real-time visibility required to monitor marketplace conditions, anticipate supply shortages, and coordinate interventions during the event’s busiest periods.
Approach

Designed and delivered a minute-level monitoring system that combined GPS telemetry, booking activity, driver availability, and marketplace performance signals. Built a unified dashboard with demand-density maps, supply-gap indicators, driver engagement metrics, and event KPIs, and set up automated Google Chat alerts.

The command center helped Operations detect emerging supply constraints, coordinate drivers, and respond to changing demand in near real time. The peak-day analysis covered approximately 2,000 users. Compared with other days’ average, the analysis reported a 10.2% increase in supply rate and a 7.4% increase in request-to-completion rate. The figures are observational.

Tools

G-Force Board: analytics app for acquisition partners

Data Director · GOZEM

Replaced support-driven status inquiries with a self-service platform serving 500+ acquisition agents
Challenge
Gozem relied on a large network of external acquisition agents to support growth across multiple markets through its G-Force program. Agents depended on Operations and Support for updates on submission status, validation progress, payment eligibility, rejection reasons, and processing delays. These requests added work for both teams and slowed acquisition operations.
Approach

Designed and built G-Force Board, a mobile-first reporting web application in Gozem Data Hub. Agents could track submissions, assessment and validation status, rejection details, payments, pending earnings, and processing timelines.

Implemented automated data synchronization, retries, and CI/CD to keep operational data available. Giving partners direct access to this information reduced their reliance on Support, helped them resolve issues faster, and supported acquisition operations across Gozem’s markets.

Tools

Data platform modernization and FinOps transformation

Head of Data · GOZEM

Reduced Google Cloud costs by 42.57% in one quarter while sustaining platform reliability and growing analytics adoption.
Challenge
As Gozem scaled across multiple countries and business units, analytics demand, reporting needs, and data volumes grew significantly. BigQuery transformation workloads relied heavily on scheduled queries with limited dependency management and visibility into cost drivers, leading to duplicated computation, increasing cloud expenditure, and growing operational complexity.
Approach

Established data platform observability and FinOps governance by analyzing BigQuery workload metadata to categorize ingestion, transformation, reporting, and ad hoc analytical activities. Identified the primary drivers of platform spend and led a modernization program spanning Dataform adoption, dependency-aware orchestration, incremental processing, partitioning and clustering optimization, reusable intermediate data models, and warehouse improvements tailored to workload access patterns.

Implemented monitoring and cost-allocation dashboards to guide ongoing optimization while maintaining data freshness and reliability as analytics usage grew.

Tools

Payments intelligence and fraud monitoring

Head of Data · GOZEM

Established Gozem Money’s analytics and risk foundations, identifying tens of millions of XOF in potential losses while supporting financial and regulatory reporting.
Challenge
As Gozem launched Gozem Money, its digital payments and money transfer business, it needed a data platform that could scale to support transaction monitoring, fraud prevention, financial reporting, and regulatory obligations. The company needed to detect external fraud attempts and internal misuse of payment processes while building reliable reporting from scratch.
Approach

Led the design and implementation of the Gozem Money data architecture across ingestion, transformation, storage, reporting, and monitoring. Built the Togo payments warehouse and ELT pipelines that supplied data for financial analytics, operational reporting, and regulatory reporting.

Designed and supervised fraud monitoring that combined rule-based screening, anomaly detection, transaction threshold monitoring, and investigative analytics. Worked with Product, Operations, and leadership to identify emerging fraud patterns, investigate suspicious activity, and turn findings into new controls, alerts, and process safeguards.

Set up an ongoing process to document newly discovered fraud schemes and operational risks and use them to update monitoring. Added data quality checks, governed metric definitions, and documented reporting logic to support reliable reporting and compliance readiness as the business grew.

Tools

Data-driven incentive optimization for delivery operations

Head of Data · GOZEM

Automated incentive management for delivery operations, improving marketplace coverage, payment processing, and dispute resolution workflows.
Challenge
As delivery volumes grew, agent coverage across high-demand zones varied, limiting marketplace efficiency and service quality. Teams had to calculate incentives, coordinate disbursements, and investigate payout disputes manually, adding work as the program grew.
Approach

Partnered with Operations and business leaders to design incentive programs around agent availability, order fulfillment, service quality, and geographic coverage. Analyzed demand by zone, time period, and operational event to identify where incentives could help match agent supply to demand.

Built a workflow that evaluated agent performance against configurable rules, calculated eligible rewards, generated payment instructions, and kept historical records for audits and dispute review. Worked with Product and Engineering to progressively integrate incentive logic into internal tools so teams could run and test new incentive programs.

Tools

E-commerce recommendation engine and personalization

Head of Data · GOZEM

Delivered a production recommendation engine. In the first month after launch, recommendations were associated with approximately 12% of delivered orders.
Challenge
As Gozem expanded its e-commerce offering, Product needed a scalable way to improve product discovery, engagement, and conversion. The team needed to choose a recommendation strategy, identify useful behavioral features, put machine learning pipelines into production, and test where to show recommendations in the customer journey.
Approach

Led the recommendation engine initiative from concept to deployment. Partnered with Data Science, ML Engineering, and Product to evaluate approaches used by leading technology platforms and choose one suited to Gozem’s customer behavior, available data, and business objectives.

Guided the design of behavioral and transactional features based on customer interactions, purchases, browsing activity, and product relationships. Oversaw the team’s implementation of feature engineering workflows, Vertex AI training pipelines, model evaluation, deployment pipelines, and serverless inference services integrated into the mobile app.

Worked with Product to define tracking events, success metrics, and the experiment design. Product used the results to select the homepage placement for rollout.

The A/B test involved approximately 600 users across three groups: no recommendations, homepage recommendations, and recommendations during search. The team observed the strongest engagement and conversion on the homepage. The 12% figure measures the share of delivered orders associated with recommendations in the first month after deployment.

Tools

Building Gozem's modern analytics foundation

Global Data Analyst · GOZEM

Established the analytical foundation that supported the company’s expansion, reducing reliance on manual reporting while enabling faster, more accessible, and more consistent decision-making across functions.
Challenge
Legacy data processes relied heavily on manual reporting and could no longer support the growing operational, analytical, and insights demands of a rapidly expanding business.
Approach
Designed and led the migration of the analytics infrastructure to Google Cloud and BigQuery, establishing a centralized data warehouse and automated data pipelines. Integrated reporting directly into tools already used across the organization, enabling self-service access to trusted data, automated dashboard refreshes, and significantly faster delivery of business insights.
Tools

Marketplace intelligence for driver retention

Global Data Analyst · GOZEM

Identified the root cause of seasonal driver churn and enabled a targeted incentive program that helped maintain marketplace participation during critical periods while creating a scalable process for payout execution and dispute resolution.
Challenge
Recurring periods of driver attrition were impacting marketplace performance. Internal operational and performance data revealed the symptom but failed to explain the underlying cause, making it difficult for business leaders to design an effective response.
Approach
Combined operational data analysis with external market research, stakeholder interviews, and contextual data collection to investigate driver behavior. Discovered that many drivers temporarily left the platform during agricultural seasons to pursue farm-related income opportunities. Translated these findings into a data-driven incentive strategy, collaborated with business leaders to validate the program design, and built the data products, allocation logic, and automation required to execute payouts at scale while ensuring full auditability.
Tools

Projects

Advanced Dataform framework

Reduced a model definition from approximately 150 lines to 20 lines of configuration in the documented example, while standardizing cost-efficient, production-ready BigQuery patterns across pipelines.
Challenge
Google Cloud Dataform simplified pipeline orchestration, but the existing setup still relied on full-table assertions, custom partition-aware MERGE logic, manual schema changes, and external orchestration for some steps. That added BigQuery scans and maintenance work as workloads grew.
Approach
Extended Google Cloud Dataform with a reusable modeling framework that standardizes advanced BigQuery engineering patterns across pipelines. The framework automates incremental processing, schema evolution, partition and clustering management, and partition-scoped data quality checks. This turned complex pipeline requirements into reusable model configurations, reducing unnecessary BigQuery scans and reliance on external orchestration.
Tools

iBola

Delivered a bilingual portfolio chatbot with automated evaluation and regression testing to continuously monitor answer quality and factual accuracy.
Challenge
Traditional portfolios require visitors to manually navigate projects, experience, and content, making it difficult to quickly find specific information or validate claims against published materials.
Approach
Designed a retrieval-augmented AI assistant powered by a LangGraph workflow that combines semantic and keyword search, grounds responses in portfolio content, validates claims against retrieved evidence, and continuously evaluates answer quality through automated regression testing and versioned benchmarks.
Tools
Welcome

Quick intro

I’m Bolaji, a Data Engineer & Architect. I design and build production data platforms and marketplace analytics solutions, and lead the teams behind them.

I design how data flows from source systems to the products teams use. I build batch and streaming pipelines, model warehouse and lakehouse data, and build governed applications.

Former Head of Data with 7+ years building data platforms. Statistics Graduate with a decade across analytics, data science, and data engineering. Hands on GCP & AWS.

Built and led a data team from 0 to 14. Delivered ~300 bigquery models and airflow dags for reporting, self-service and operational tools. Led workloads optimisation that reduced GCP cost by 42.57% in one quarter.