Open to Senior & Lead AI / Agentic AI roles
I build AI systems that are useful, measurable, and ready for production.
I design intelligent systems across machine learning, LLMs, agentic workflows, evaluation and infrastructure — with a bias toward reliability, cost, latency and outcomes that a business can actually point at.
- specialised agents in production
- 12specialised agents in production
- schema-validated tools
- 280+schema-validated tools
- evaluation suite score
- 95%+evaluation suite score
AI systems with intent.
$ ship --reliable-ai
- ✓ evaluation gate passed
- ✓ latency optimised
- ✓ safeguards enabled
Selected work
Work that goes beyond the demo.
Systems where the model was the easy part. Each one opens a case study covering the architecture, what I built and what it changed.
Earlier work
Systems thinking
I care about the layer beneath the model.
Reliable AI is a loop: retrieve the right context, generate with guardrails, evaluate the result, observe what actually happened, then improve the system on purpose rather than by feel.
- 01
Evaluate
Golden datasets, multi-turn tests, a version-pinned judge rubric, and quality gates that block a bad release.
- 02
Observe
Traces across prompts, retrieval, tool calls, latency, cost and outcomes - with the feedback loop closed.
- 03
Operate
Tool boundaries, validators between proposal and execution, access control, and the safeguards that keep a system upright under real traffic.
Engineering principles
The way I approach AI engineering.
Six things I keep coming back to, whatever the model or the framework happens to be that quarter.
Make it measurable
Define what good looks like before optimising anything. Evaluation is part of the product, not a phase after it.
Never let the model hold the keys
An LLM proposes; a validator decides; an executor acts. Money paths and destructive actions stay deterministic, so a bad generation is a bad sentence rather than a bad transaction.
Ground every claim
A figure the system states must trace back to something a tool actually returned. When it cannot, fail closed rather than ship a confident lie.
Optimise the whole path
Model choice is one lever. Caching, routing, streaming and orchestration shape cost and latency just as much.
Protect the data boundary
AI systems should reach business data through controlled, auditable interfaces - never unrestricted database access.
Ship for reality
Containers, cloud infrastructure, monitoring and failure handling are what separate a demo from a system.
ABOUT SAMUEL
From models to systems.
I started in machine learning. Today I build AI systems that have to survive contact with real users, real data and real money.
I'm Lead AI/ML Engineer at BrandDrive, where I own a production multi-agent platform sitting in front of a banking product. Before that: four years of classical ML — forecasting, classification, feature engineering — which is still the foundation everything else rests on.
What changed is the stakes. Most of my work now involves agents with real authority: over a bank account, over a business's books. That turns the interesting questions from "can the model do it?" into "what happens when it gets it wrong, and who is holding the keys when it does?"
The part I enjoy most is what comes after the first successful demo: how do we measure it, how do we make it cheaper, how do we make it safer, and how do we keep it working?
- Based in
- Abuja, NG
- Focus
- Agentic AI + ML
- Experience
- 4+ years
Experience
Where the work has happened.
Four years across production AI, consulting and teaching — the last of which is where I learned to explain a system before defending it.
Mar 2025 — present
Abuja, Nigeria
Lead AI/ML Engineer · BrandDrive
- Architected and deployed a production multi-agent AI platform for business and personal banking across web, mobile and WhatsApp.
- Built the LLM evaluation and regression-testing framework that moved system quality from ~55% to consistently above 95%.
- Engineered cost-aware model routing down to ~$0.007 per request, and cut latency through prompt caching, incremental streaming, agent reuse and fast-path execution.
- Designed secure execution for sensitive financial operations: human-in-the-loop approval, permission-aware tool access, auth propagation and PIN verification.
- Python
- TypeScript
- LangChain
- LangGraph
- Node.js
- AWS
Nov 2023 — Feb 2025
Remote
Machine Learning Consultant · Independent
- Delivered end-to-end ML solutions for clients in fintech, e-commerce and healthcare — data preparation through to deployment and business integration.
- Built predictive analytics for forecasting, classification and decision support using supervised and deep learning methods.
- Translated business requirements into production-oriented ML, defining objectives and evaluation metrics with stakeholders rather than inheriting them.
- Python
- TensorFlow
- scikit-learn
- Pandas
- SQL
- Tableau
Mar 2022 — Nov 2023
Minna, Nigeria
Chief Data Science · CADEMIT
- Led the data science training programme, mentoring 30+ aspiring data scientists through the full ML lifecycle.
- Designed and delivered 120+ hours of hands-on training and 15+ technical workshops across ML, deep learning, NLP, Python, SQL and Git.
- Built project-based curricula focused on turning theory into engineering that actually ships.
- Python
- SQL
- TensorFlow
- scikit-learn
- Jupyter
- Git
- Education
- BSc Computer ScienceFederal University of Technology, Minna
- Speaking
- Keynote, Nigerian Society of Engineers“The Synergy of AI and Engineering” — 100+ engineers, Nov 2023
- Mentoring
- 30+ data scientists trained120+ hours of instruction, 15+ workshops delivered
Toolkit
Comfortable across the stack.
The stack is a means to an outcome. I reach for whatever makes the system dependable and maintainable, and I'm happy to learn the rest.
LLM & Agents
- LangChain
- LangGraph
- Tool Calling
- RAG
- OpenAI
- Anthropic
- Gemini
- AWS Bedrock
- Groq
- Transformers
Evaluation & Reliability
- Golden Datasets
- LLM-as-Judge
- Regression Testing
- Guardrails
- Observability
- Audit Trails
- Jest / Pytest
Machine Learning
- Python
- scikit-learn
- TensorFlow
- Keras
- XGBoost
- LightGBM
- Time Series
- NLP
Production & Infra
- AWS
- Docker
- Kubernetes
- CI/CD
- TypeScript
- Java 17
- NestJS
- FastAPI
- PostgreSQL
- React / Next.js
Contact
Let’s build something that deserves to be shipped.
Actively looking for Senior or Lead AI/ML Engineer and Agentic AI Engineer roles — production AI, multi-agent systems, evaluation and reliability. Based in Abuja, Nigeria, working remotely with teams anywhere.