Oluwagbade OdimayoData and Cloud Engineer

Data and Cloud Engineer

I build reliable data platforms, cloud infrastructure and reproducible workflows for operational and scientific data.

I use Python, SQL and cloud-native infrastructure to build testable, observable and reproducible systems for operational and scientific data.

View selected workRésumé available on request

  • Data Engineering

    Reliable ingestion, transformation, modelling, validation, reconciliation and monitoring.

  • Cloud and Platform Engineering

    Infrastructure as code, containers, Kubernetes, CI/CD and deployment automation.

  • Bioinformatics and Scientific Workflows

    Reproducible pipelines, provenance, data integration and research-software engineering.

  • Applied AI and Machine Learning

    Leakage-aware modelling, calibration, anomaly detection and decision systems built around real operating constraints.

Supported by research publications, open-source contributions and technical writing.

Selected engineering work

Systems designed to be tested, operated and understood

A focused selection of data, cloud and scientific-computing systems with documented architecture, validation, limitations and operational evidence.

01

Data Engineering

AWS SupplyShield Lakehouse

Stops unreliable supply-chain data reaching the people making purchasing and inventory decisions.

A cloud data platform that ingests operational supply-chain data incrementally, validates its quality and produces reconciled datasets for downstream analytics.

data flowOperationalsourcesIncrementalingestionRaw zoneAmazon S3Contracts +schema checksModelledPostgreSQLSource-to-targetreconciliationQuarantine + alertmonitoring · structured logging · freshness checks
SupplyShield: incremental ingestion, contract enforcement, reconciliation and monitoring

Records land in an Amazon S3 raw zone before passing contract and schema validation. Anything failing validation is quarantined and alerted rather than silently dropped. Modelled output is reconciled source-to-target so the numbers can be defended, and the whole path is provisioned by Terraform and validated in CI.

02

Cloud and Platform Engineering

Enterprise Progressive Delivery Platform

Lets a team ship often without betting the service on every release. A bad deployment rolls back in seconds rather than becoming an incident.

A Kubernetes release platform that validates application health before promotion and supports controlled rollback when deployment conditions fail.

data flowgit pushmainCI scansTrivy · CheckovTerraformAWS + EKSArgo RolloutscanaryHealth gateevaluationPromoteto stableAutomated rollbackkubernetes · nginx ingress · teardown and cost control
Progressive delivery: canary evaluation with automated rollback on Amazon EKS

Terraform provisions AWS and Amazon EKS; Argo Rollouts releases a canary that is evaluated against health gates before promotion. The case study documents rollback validation and an unsuccessful rollout state as engineering evidence rather than hiding it.

03

Bioinformatics and Scientific Computing

ECM Target Discovery Pipeline

Narrows a large candidate space to a reviewable shortlist, with the evidence behind every candidate traceable.

A reproducible workflow that integrates biological evidence, records provenance and produces reviewable candidate-prioritisation outputs.

data flowPublicdatabasesNextflowcontainerisedQualitycontrolEvidenceintegrationCandidatescoringReviewablereportQC failure halts runprovenance manifest · conda · metadata capture · reproducible environments
ECM target discovery: provenance-preserving evidence integration

Public biological databases feed a containerised Nextflow workflow. Quality-control failure halts the run rather than producing partial results, and every output ships with a provenance manifest recording each input, version and transformation.

The workflow supports research prioritisation. It does not replace biological interpretation, experimental validation or clinical review.

04

Applied Machine Learning

Transaction Fraud Scoring Platform

Catches more fraud without hiring more analysts, by ranking cases against the review capacity that actually exists.

A batch fraud-prioritisation system that combines calibrated risk scores, business rules and operational capacity constraints.

data flowTransactionstreamPoint-in-timefeaturesModel +calibrationBusinessrulesCapacity-awarequeueAnalystreviewLeakage checksdrift monitoring · batch scoring · pytest · reproducible experiments
Fraud scoring: leakage-aware features feeding a capacity-bounded review queue

Features are built point-in-time correct with explicit leakage checks. The review queue is bounded by real analyst capacity rather than a fixed probability threshold, and analyst outcomes feed drift monitoring.

How I build

Principles that shape my work

  • Reproducibility

    Infrastructure, dependencies and processing steps should be defined in code.

  • Failure-aware design

    Retries, idempotency, rollback and recovery must be designed before failure occurs.

  • Observability

    Systems should expose enough information to identify what happened, where it happened and what data was affected.

  • Honest scope

    Portfolio systems must not be described as enterprise production systems unless they have genuinely served users under operational responsibility.

Expertise

Engineering expertise

  • Data Engineering

    • Incremental ingestion
    • Data modelling
    • Data quality
    • Reconciliation
    • Orchestration
    • Observability

  • Cloud and Platform Engineering

    • AWS and Azure
    • Terraform
    • Docker
    • Kubernetes
    • CI/CD
    • Deployment validation

  • Bioinformatics and Scientific Computing

    • Workflow orchestration
    • Biological data integration
    • Provenance
    • Reproducible environments
    • Quality control
    • Research-software documentation

  • Applied AI and Machine Learning

    • Leakage prevention
    • Feature engineering
    • Classification
    • Calibration
    • Drift monitoring
    • Decision-system integration
Microsoft credentials

Certified across Microsoft Fabric data engineering and analytics

Role-based credentials covering the engineering and analytics layers of Microsoft Fabric.

  • DP-700

    Microsoft Certified

    Microsoft Certified: Fabric Data Engineer Associate

    Data loading patterns, data architectures, orchestration, transformation, security, monitoring and optimisation in Microsoft Fabric.

    Verify on Microsoft Learn ↗

  • DP-600

    Microsoft Certified

    Microsoft Certified: Fabric Analytics Engineer Associate

    Enterprise analytics solutions, lakehouses, warehouses, semantic models, data preparation, security and analytics lifecycle management in Microsoft Fabric.

    Verify on Microsoft Learn ↗
Selected research

Research that strengthens my engineering practice

  • 2025 · Journal article

    Exploration of Plant Alkaloids as Potential Inhibitors of HIV–CD4 Binding

    Computational biology · Drug discovery

  • 2025 · Conference / preprint

    Temporal-Spatial Attention Network for DoS Attack Detection

    Machine learning · Cybersecurity

  • 2024 · Journal article

    DNA Methylation in the Recovery of Maize from Maize Streak

    Molecular genetics · Plant biology

Open-source engineering

Contributing to systems other people maintain

Understanding an existing architecture, following maintainer direction, reproducing defects, writing focused tests and accepting review. Code merged upstream into Meltano and nf-core/modules.

  • Meltano ↗

    3 merged PRs

    Three merged contributions to the open-source DataOps and ELT platform: suppressed coloured output when stderr is not a TTY so redirected logs stay clean, added FORCE_COLOR support with NO_COLOR precedence and test coverage, and added --refresh-catalog to meltano invoke for Singer tap discovery.

    Python · CLI design · logging · Singer catalogue handling · tests · documentation

    Merged

  • nf-core/modules ↗

    PR #12018

    Fixed the FreeBayes module. Three nf-test cases ran on human data but were labelled sarscov2, so they were renamed and their snapshot keys updated. Added the missing EDAM ontology entries for BAM, CRAM, SAM, BED and FASTA inputs, and corrected stub gzip syntax required by nf-core linting.

    Nextflow · nf-core · nf-test · EDAM ontology · YAML · CI

    Merged

  • Pydantic Settings ↗

    Developed regression tests for nested environment-variable aliases, validation aliases and alias choices, with pinned environments and failure analysis.

    Python · Pydantic · pytest · environment management

    Local validation

  • AEGIS ↗

    Replaced inaccurate installation instructions with a reproducible local setup for an AI-governance application.

    Technical documentation · reproducibility · Streamlit

    Awaiting maintainer action

Selected writing

Explaining how systems are built, tested and improved

  • Building an Incremental Supply-Chain Lakehouse on AWSPlanned
  • What I Learned from an Argo Rollouts FailurePlanned
  • Applying Data Engineering Principles to Bioinformatics WorkflowsPlanned

I publish on

About

Building dependable systems for data-intensive work

My background combines applied data science, cloud engineering and life sciences. I build dependable systems for operational and scientific data, with particular strength in data platforms, infrastructure automation, bioinformatics workflows and reproducible computation.

Contact

Open to data, cloud and scientific computing opportunities

Data engineering, cloud platforms, platform engineering and bioinformatics workflow roles.

hello@gbadedata.com