Senior AI Engineer · Portfolio
Ovo Okpubuluku
Technical Lead, Production AI Systems
Building production AI systems for Canada's largest financial institutions. From banking operations in Nigeria to safety-critical field crews in Alberta to Senior AI Engineer at a tier-1 bank. The industries changed. The discipline didn't.
01 — Career Arc
Twenty years.
Three industries. One discipline.
The arc runs from branch banking operations in Nigeria, through safety-critical field leadership in Alberta's energy sector, into data science and production AI engineering in Canada. The industries changed. The pressure, the accountability, and the requirement to deliver without excuses did not. Leadership under consequence is not a credential you earn in a classroom. It is a track record you build across industries.
Senior AI Engineer
Tactable · Remote, Canada
Technical Lead on production AI systems for a tier-1 Canadian bank and global AdTech clients. Owns end-to-end delivery from architecture through deployment — RAG agents, agentic workflows, ML inference pipelines, and a full-stack AI campaign management platform. Manages everything the audit team touches.
Senior Data Engineer
Badal by TELUS Digital · Remote, Canada
Led sprint execution and delivery for enterprise data engineering teams at Canada's foremost GCP consulting firm. Drove the migration of a complex on-prem Hadoop Data Lake to GCP and built production agentic AI workflows for Regulatory Reporting and Customer Service at tier-1 banking scale. Built secure GenAI architecture using privately hosted Vertex AI endpoints.
Applied ML Engineer -> Data Scientist/Engineer
ReVisionz Inc. → Totogi LLC · Calgary / Remote
Delivered computer vision solutions: YOLO Object Detection and Azure Cognitive Services OCR, achieving 50%+ cost reduction for Suncor Energy's Asset Data Lake Platform. Then led technical delivery of the Totogi-AWS Pilot Project: a first-of-kind 4G/5G VoLTE Network on AWS using Magma Core Architecture.
Plant Lead & Crew Lead
United Safety / Safety BOSS / Precise Downhole · Alberta
Led safety-critical field operations teams for Suncor Energy, Shell Canada, MEG Energy, and Enbridge. Co-Chaired the Joint Health and Safety Committee at Precise Downhole Solutions. The discipline of leading crews under genuine consequence is the same discipline I apply to production AI systems under load.
Head of Branch Operations
First City Monument Bank (FCMB) · Asaba & Ekpoma, Nigeria
Led full branch operations for FCMB across two locations: staff, contractors, enterprise risk, vault, and profitability. Designated organizational change agent for a bank-wide transformation initiative. Zero-fraud record across the entire tenure.
02 — Production Work
Systems shipped.
In production. Under audit.
Enterprise RAG Agent
Tier-1 Canadian Bank · Microsoft Azure Stack
RAG Agent answering questions across millions of documents. Feature development, production improvements, and end-to-end delivery through security review, compliance, and enterprise audit at tier-1 banking scale.
Azure AI Foundry, Azure OpenAI, Qdrant, LangGraph, Python, FastAPI
AI Campaign Management Platform
Global AdTech Platform · Digital Out-of-Home Advertising
End-to-end delivery of a full-stack AI application for DOOH campaign analysis and booking — solution architecture through production deployment. Owned the entire delivery lifecycle.
Azure Container Apps, Bicep IaC, LangGraph, Python, React
AdTech ML Inference & Retraining Pipeline
AdTech Platform · Creatives Classification at Scale
Production maintenance, retraining pipelines, and platform enhancements for a classical ML and AI classification inference system. ONNX Runtime for edge inference with automated retraining on data drift.
ONNX Runtime, Azure ML, MLOps, Python, Docker
Enterprise Data Lake Modernization
Bank of Nova Scotia · Global Data Analytics Platform
Designed and deployed a CI/CD analytics pipeline migrating a complex on-prem Hadoop Data Lake and Warehouse to GCP. Hive/Spark ETL, Qlik ingestions, SFTP pipelines, TAMR datasets at full banking scale.
BigQuery, dbt, Cloud Composer, Dataproc, Terraform
Agentic AI Workflow Suite
Tier-1 Bank · Regulatory Reporting + Customer Ops
Production-grade agentic AI workflows for Regulatory Reporting, Customer Service, and domain research. LangGraph multi-agent orchestration with HITL oversight. Secure GenAI architecture with privately hosted Vertex AI.
LangGraph, LangSmith, Qdrant, Vertex AI, RAG
Asset Data Lake Computer Vision
Suncor Energy · Asset Data Lake Platform
50%+ Cost Reduction
Delivered AI/ML solutions achieving 50%+ cost reduction on Suncor Energy's ADLP using YOLO Object Detection and Azure Cognitive Services OCR on industrial asset data at scale.
YOLO, Azure Cognitive, Python, Computer Vision, Azure
03 — Built in public
Public repositories.
Client work stays under NDA. These are the public builds: the same patterns, in code you can read.
AI engineering
InvestigatorAI
A LangGraph team of agents for fraud analysts: regulatory research, evidence collection, a compliance check, then a filing-ready report. Retrieval is hybrid, BM25 plus dense vectors in Qdrant with Cohere reranking, scored with RAGAS.
LangGraph, FastAPI, OpenAI, Qdrant, LangChain, scikit-learn, Redis, Tavily, LangSmith, Next.js, React, TypeScript, Tailwind CSS, Python, pandas, PyMuPDF, Docker Compose, Docker, GitHub Actions, pytest, uv, Railway, Vercel
EvolSynth API
Builds evaluation sets from your own documents with Evol-Instruct: simple questions evolved into multi-step reasoning questions, then graded by an LLM judge. It's the test data a RAG system needs before anyone trusts its scores.
FastAPI, LangChain, OpenAI, Redis, LangSmith, Next.js, React, TypeScript, Tailwind CSS, Python, PyMuPDF, Docker Compose, Docker, pytest, Railway, Vercel
MCP Server App
A Model Context Protocol server with custom tools, driven by a LangGraph ReAct agent through langchain-mcp-adapters, with its own pytest suite.
Model Context Protocol, LangGraph, LangChain MCP Adapters, OpenAI, FastAPI, Tavily, Next.js, React, TypeScript, Python, GitHub REST API, Unsplash API, Docker Compose, Docker, pytest, uv
ReplySight
A support agent pastes a complaint, and a LangGraph agent on GPT-4o calls Tavily web search before GPT-4o-mini drafts a reply with citations. FastAPI serves the graph to a Next.js UI, with offline pytest checks in CI.
LangGraph, OpenAI, LangChain, FastAPI, Tavily, LangSmith, Next.js, React, TypeScript, Tailwind CSS, Python, Docker Compose, Docker, GitHub Actions, pytest, Railway, Vercel
PostAssist
A LangGraph supervisor runs a content team and a verification team to turn an ML paper title into a fact-checked LinkedIn post. FastAPI runs each job in the background with Redis task state, polled by a Next.js dashboard.
LangGraph, LangChain, OpenAI, FastAPI, Tavily, Redis, Next.js, React, TypeScript, Tailwind CSS, Python, Docker Compose, Docker, GitHub Actions, pytest, Railway, Vercel
Resume Parser
PDF and Word resumes parsed with PyMuPDF and python-docx, names found with spaCy NER, and skills pulled by GPT in JSON mode or an offline keyword matcher. Served through FastAPI and a React upload UI, with unit and integration tests.
OpenAI, spaCy, FastAPI, PyMuPDF, python-docx, React, TypeScript, Vite, Python, Nginx, Docker Compose, Docker, GitHub Actions, pytest, uv
Machine learning
Income Prediction Pipeline
Train, track, serve, watch: XGBoost and scikit-learn models tracked in MLflow, orchestrated with Prefect, served through Flask in Docker, with Evidently drift reports. My most-starred repo.
XGBoost, MLflow, Prefect, Evidently, Flask, Prometheus, Grafana, MongoDB, Streamlit, scikit-learn, Hyperopt, imbalanced-learn, pandas, Apache Parquet, Docker Compose, Docker, Gunicorn, Pipenv, pytest, pre-commit, Python
House Price Model Pipeline
From research notebook to production package: a scikit-learn gradient-boosting model with its own feature transforms and test suite, trained on first boot and served through FastAPI to a Streamlit form, run with Docker Compose.
scikit-learn, FastAPI, Streamlit, pandas, SciPy, Pydantic, Docker Compose, Docker, pytest, Python
Car Manufacturing Test
Cutting the time a Mercedes-Benz car spends on the test bench: dimensionality reduction across the build features, then an XGBoost model.
XGBoost, scikit-learn, FastAPI, Streamlit, pandas, Pydantic, Docker Compose, Docker, pytest, Python
Salary Prediction
A HistGradientBoosting model trained on 1M job postings during docker build, served through a FastAPI batch endpoint to a React explorer, with target leakage found in the original notebook documented and fixed.
scikit-learn, FastAPI, React, TypeScript, Vite, pandas, Pydantic, Nginx, Docker Compose, Docker, GitHub Actions, uv, pytest, Python
Customer Segmentation
UCI Online Retail transactions scored into RFM segments and K-Means clusters at docker build, with cohort retention, served by FastAPI to a React dashboard that assigns new customers to a segment.
scikit-learn, pandas, FastAPI, React, TypeScript, Vite, Pydantic, Nginx, Docker Compose, Docker, GitHub Actions, uv, pytest, Python
Employee Attrition Predictor
A feature-engine and scikit-learn pipeline trained in a compose job that logs every run to MLflow, tags the best by ROC AUC as champion, and serves that registry alias through FastAPI to a React app.
MLflow, scikit-learn, Feature-engine, FastAPI, React, TypeScript, Vite, pandas, Pydantic, Nginx, Docker Compose, Docker, GitHub Actions, uv, pytest, SQLite, Python
Data engineering
AWS ETL Pipeline
DynamoDB exports landed in S3 and transformed by Glue jobs on Spark, then queried with Athena from Lambda functions, with a local DuckDB replay and a Streamlit dashboard under Docker Compose.
AWS Glue, Amazon Athena, AWS Lambda, Amazon S3, Apache Spark, Amazon DynamoDB, Python, SQL, DuckDB, Streamlit, pandas, Apache Parquet, Altair, Docker Compose, Docker, pytest
Analytics Engineering Prototype
A Kimball dimensional model in dbt on BigQuery, for a fictional specialty-foods importer moving its reporting off the operational database.
dbt, PostgreSQL, BigQuery, SQL, Streamlit, Plotly, SQLAlchemy, pandas, Python, Docker Compose, Docker
Document Streaming Pipeline
JSON events through FastAPI into Kafka, processed with Spark Structured Streaming, stored in MongoDB and read in a Streamlit dashboard.
Apache Kafka, Spark Structured Streaming, MongoDB, FastAPI, Streamlit, Python, Pydantic, pandas, Jupyter, Docker Compose, Apache ZooKeeper, Docker
Airflow ETL Pipeline
Apache Airflow pulls books from Open Library and daily stock bars from Yahoo Finance into a parquet data lake, then loads a Postgres warehouse, with a FastAPI and React viewer over each run.
Apache Airflow, PostgreSQL, pandas, yfinance, Open Library API, FastAPI, React, TypeScript, Python, SQLAlchemy, Apache Parquet, Docker Compose, GitHub Actions, pytest, Vite, Docker, Nginx, uv
Time-Series Analytics
Air-quality and live weather series written to InfluxDB, queried with Flux, shown in auto-provisioned Grafana dashboards, with a FastAPI and React console that flags rolling z-score anomalies.
InfluxDB, Flux, Grafana, FastAPI, React, Terraform, Cloud Run, Azure Container Apps, Amazon ECS on Fargate, Python, TypeScript, pandas, Open-Meteo API, WeatherAPI, Docker Compose, GitHub Actions, pytest, Docker, Vite, Artifact Registry, Azure Container Registry, Amazon ECR, Secret Manager, AWS Secrets Manager, TFLint, Checkov, Nginx, uv
Contact Tracer
Synthetic app check-ins at San Francisco businesses indexed in Elasticsearch, where term and time-range queries find who shared a place and time window, served by FastAPI to a React and Leaflet map with geo-distance business lookup.
Elasticsearch, FastAPI, React, Leaflet, Faker, Terraform, Cloud Run, Azure Container Apps, Amazon ECS on Fargate, Python, TypeScript, OpenStreetMap, Docker Compose, GitHub Actions, pytest, Docker, Vite, Artifact Registry, Azure Container Registry, Amazon ECR, Secret Manager, AWS Secrets Manager, TFLint, Checkov, Kibana, Nginx, uv
Cloud & DevOps
Python Azure AI REST APIs
FastAPI services for Azure AI Language sentiment and key phrases plus Tesseract OCR, tested and deployed to Azure App Service by GitHub Actions, with a Streamlit front end.
FastAPI, Python, Azure AI Language, Tesseract OCR, Azure App Service, Streamlit, Pydantic, Docker Compose, Docker, GitHub Actions, pytest
GCP Terraform App Deployment
Terraform for a remote state bucket, an Artifact Registry image and a Cloud Run service.
Terraform, Cloud Run, GKE, Artifact Registry, Django, Python, Docker, Cloud Storage, SQLite
Azure Streaming Pipeline
Tweets posted to a FastAPI endpoint are schema-checked, landed in Blob Storage and streamed through Event Hubs into Cosmos DB, running locally on Azurite and the Azure emulators with a React live view.
Azure Functions, Azure Event Hubs, Azure Cosmos DB, Azure Blob Storage, Azurite, FastAPI, Python, React, TypeScript, Server-Sent Events, JSON Schema, Docker Compose, Docker, Vite, Nginx, uv, GitHub Actions, pytest
Multi-cloud Terraform Modules
A tagged library of 19 Terraform modules for Cloud Run, Azure Container Apps and AWS ECS Express Mode, with Postgres, registries and remote-state backends for each cloud. CI runs fmt, validate, tflint and checkov.
Terraform, Cloud Run, Azure Container Apps, Amazon ECS on Fargate, GKE, PostgreSQL, Cloud SQL, Azure Database for PostgreSQL, Amazon RDS, Artifact Registry, Azure Container Registry, Amazon ECR, Memorystore for Redis, Secret Manager, AWS Secrets Manager, Cloud Storage, Amazon S3, Amazon DynamoDB, Azure Storage, Azure Monitor Log Analytics, Amazon CloudWatch, GitHub Actions, TFLint, Checkov
US City API
One Terraform root per cloud deploys the Express API and React map to Cloud Run, Azure Container Apps or ECS Express Mode, each with a managed Postgres 16 database, using shared modules pinned to a git tag.
Express, Node.js, PostgreSQL, JavaScript, SQL, Terraform, Cloud Run, Azure Container Apps, Amazon ECS on Fargate, Cloud SQL, Azure Database for PostgreSQL, Amazon RDS, React, TypeScript, Leaflet, OpenStreetMap, Vite, Nginx, Docker Compose, Docker, Jest, Supertest, Artifact Registry, Azure Container Registry, Amazon ECR, Secret Manager, AWS Secrets Manager, GitHub Actions, TFLint, Checkov
Java Spark Batch Application
Terraform runs the Spark batch as a Cloud Run job, Container Apps job or Fargate task that writes to Cloud SQL, Postgres Flexible Server or RDS, with a FastAPI run viewer and React front end beside it.
Apache Spark, Java, Spring Boot, PostgreSQL, SQL, Terraform, Cloud Run, Azure Container Apps, Amazon ECS on Fargate, Cloud SQL, Azure Database for PostgreSQL, Amazon RDS, FastAPI, Python, React, TypeScript, Maven, H2, JUnit 5, Docker Compose, Docker, Vite, Nginx, Artifact Registry, Azure Container Registry, Amazon ECR, Secret Manager, AWS Secrets Manager, GitHub Actions, pytest, uv, TFLint, Checkov
40+ certifications on Credly (opens in a new tab)Credential wallet on Accredible (opens in a new tab)
04 — Stack & Tools
What I build with.
- Azure AI Foundry
- Azure OpenAI
- LangGraph
- LangSmith
- LangChain
- Hugging Face
- Qdrant
- ONNX Runtime
- Azure ML
- Vertex AI
- Databricks
- Docker
- Spec-driven development
- Claude Code
- Cursor
- Model Context Protocol
- Microsoft Azure
- Azure Container Apps
- Bicep
- Google Cloud
- Terraform
- AWS
- dbt
- BigQuery
- DuckDB
- ADLS Gen2
- Apache Airflow
- Apache Spark
- Python
- TypeScript
- FastAPI
- React
- Next.js
Certifications & qualifications
- Google Cloud Professional ML Engineer
- AWS Machine Learning Specialty
- Databricks Machine Learning Associate
- Azure Data Scientist Associate
- Azure AI Engineer Associate
- AI Makerspace Certified AI Engineer
- Google Cloud Professional Data Engineer
- Google Cloud Associate Cloud Engineer
- Databricks Data Engineer Professional
- HashiCorp Terraform Associate
- Quantellia Certified Decision Intelligence Analyst
- Digital MBA, Technology Leadership (CTO Academy)
- MBA, Startup Ecosystems, in progress (Augment Business School)
05 — Teaching & Mentorship
Building people.
Not just products.
AI Makerspace
Lead breakout sessions on production LLM development, RAG systems, and agentic workflows. Mentor learners on product ideation and production AI implementation. Push from what the bootcamp teaches to what production actually demands.
- LLMs in Prod
- RAG Systems
- Agentic AI
Lighthouse Labs
Delivered expert content across the Data Bootcamp and corporate GenAI training. Conducted code reviews, supported curriculum development, and guided interns from theory to production thinking over a three-year engagement.
- Data Science
- Analytics
- GenAI
Digital Craftsmanship
Production patterns from the build floor at Canada's largest banks. RAG agents, agentic workflows, enterprise cloud. No demos. No hype. Translated into frameworks engineers can use. Launching soon.
- Newsletter
- Enterprise AI
- Coming Soon
Built for Monday morning, not for the demo.
One 90-minute working session, then a written gap map and your top 3 decisions.