About
I'm an AI engineer, but most of what I build is the system an LLM application actually runs on: retrieval over vector stores with LangChain, orchestration between models and the tools they call, and the backend holding it up, services behind FastAPI, queues, auth, and databases. Machine learning projects run alongside that, mostly computer vision and NLP that has to reach production rather than stop at a notebook.
My background as a medical laboratory scientist gave me a deep appreciation for precision, rigour, and the real consequences of a result being wrong. What I care about most is what happens after that: latency, cost, and whether it still behaves once real traffic reaches it.
Beyond the technical work, that same instinct shows up in how I report results: a system that tells you not just an answer, but how confident it is and why, is often more useful to a colleague than a cleverer model underneath it.
Currently taking on reporting automation and AI engineering projects for agencies and data-heavy teams.
Experience
-
Apr 2026 – Oct 2026
AI & Automation Engineer · Welcome Tomorrow
Contract
Marketing analytics and AI reporting platform
- Built an end to end analytics platform on GCP, ingesting advertising, search and analytics APIs for agency clients on an automated schedule.
- Designed an idempotent ETL pipeline and a layered dbt project, fixing production data integrity bugs including pagination truncation, schema drift and a metric weighting error in KPI calculations.
- Architected a LangGraph multi-agent layer over the warehouse for analyst Q&A, with separate stages for intent checking, text-to-SQL and validation, each independently verifiable.
- Added guardrails around LLM generated queries: per-client data isolation, AST based SQL validation with sqlglot, session scoped access controls and an adversarial eval suite in LangSmith.
- Built a slide reporting engine exposed as an MCP server so AI can generate reports with no bespoke backend.
- Self-hosted n8n on GCP with Docker Compose, PostgreSQL and nginx, wiring website forms, a CRM, calendar bookings and team alerts through webhooks.
- Built a real-time alerting system with the analytics team to flag anomalies in performance data, scoring each day against the same day the previous week.
Meeting intelligence pipeline
- Built a production platform on FastAPI and serverless containers that auto-records meetings, summarises them and assigns action items across team calendars.
- Used verified webhooks and a message queue to decouple ingestion from AI processing, keeping the request path fast.
- Extracted tasks from transcripts with an LLM and routed them to a central task database and team messaging, with task DMs and urgency based reminders.
- Made the pipeline fault tolerant and idempotent with atomic database locks, retry caps and checkpointing, observable through structured logging and monitoring metrics.
- Shipped with 100+ unit tests against mocked external services, run in CI.
-
Sep 2025 – Feb 2026
AI/ML Engineer · Freelance
- Completed two short engagements under confidentiality agreements, so client details are withheld.
- Implemented retrieval-augmented generation systems using semantic search over vector embeddings.
- Built recommendation systems using similarity based ranking.
-
Dec 2023 – Jul 2025
Quality Assurance Officer · Cerba-Lancet Laboratories Nigeria
- Built a Python and Pandas pipeline to clean and transform raw laboratory datasets, automating visualisation and surfacing insights for weekly strategic decisions.
- Developed an automated notification system with smtplib to flag upcoming document expirations, improving compliance and readiness.
- Analysed quality metrics such as turnaround time and trends to find patterns and support process improvements.
- Managed structured documentation for data integrity and version control.
Projects
-
PlasmoScan
Malaria parasite detection
A fine tuned YOLOv11n model exported to ONNX for fast CPU inference, served from a FastAPI backend with async inference and Prometheus and Grafana monitoring, behind a React frontend with live detection overlays. Containerised and deployed across Render and Vercel with automated model fetching and CI. I ran the data end to end, annotating microscopy images in LabelImg and tracking every experiment in Weights & Biases.
-
ISO 15189 QMS Assistant
Hybrid RAG over a laboratory standard
A grounded assistant over the ISO 15189:2022 standard that answers with page level citations rather than from model memory. Retrieval runs dense pgvector search and BM25 in parallel, fuses them with Reciprocal Rank Fusion and re-ranks with a cross encoder, while an upstream LLM call rewrites follow ups into standalone questions and routes them by intent. It runs on AWS Lambda container images, a streaming FastAPI function and an SQS ingestion worker, with a Next.js frontend streaming over SSE. With no benchmark for a corpus this specific, I built an evaluation harness measuring faithfulness, retrieval Hit Rate and MRR, and abstention on out of domain questions.
-
AnimeMatch
Content based recommender
A recommender over anime metadata, building TF-IDF embeddings from synopsis, genre and theme, then ranking by cosine similarity. The dataset refreshes itself weekly from the Jikan API on a GitHub Actions schedule, recomputing the similarity matrix and pulling trending titles. Deployed on Render with embeddings persisted on Hugging Face.
-
World Mortality Analysis
SQL and Tableau
An exploratory analysis of global mortality data, querying and aggregating in SQL, reshaping with Power Query, and building an interactive Tableau dashboard. It surfaced the leading causes of death globally and in Nigeria, including a clear rise in deaths from drug use disorders.
-
Nashville Housing Cleanup
SQL transformation
End to end cleaning of a real estate dataset in SQL. Standardising dates, recovering missing addresses through self joins, splitting composite address fields, normalising categoricals, and removing duplicates with CTEs and window functions.
Skills
LLM & AI
Backend Engineering
Databases
Data Engineering
Machine Learning
Cloud & Deployment
Observability & Interfaces
Certifications
Contact
Available for freelance AI engineering and automation projects, and open to full-time roles. Email is the fastest way to reach me.