ML Platform Architect & Technical Lead · Cape Town, available globally

Signal from noise.
Architecture that ships.

I define the infrastructure and delivery strategy for production ML systems — the layer that determines whether enterprise AI creates commercial value or stays in notebooks.

Explore work Get in touch Chat with my CV

$ andrea --chat

> Ask your own question ↗

£40M+

First-year ROI from platform optimisation I led.

50+ / day

Models trained daily, up from ~1 per week, via Prediction Factory I conceptualised.

~90%

Reduction in P1/P2 incidents through MLOps standardisation.

70%

Faster experiment turnaround through champion model architectures.

Third-party perspective

What people I've worked with say

"Andrea operates with the strategic mindset of a Lead Architect. He possesses the rare ability to bridge the gap between 'Scientific Rigour' and 'Operational Speed'."

Chief Product & Prediction Officer — JUMO

"Andrea is a standout technical leader within our practice. His work consistently sets the standard for technical excellence, and he operates as a force multiplier; his presence raises the technical bar for the entire team."

Business Unit Lead & Delivery Lead — Equal Experts

"Andrea did not simply teach me technical skills; he taught me how to think like an architect. Because of his mentorship, I was promoted to AI Adoption Lead — a transition that typically takes two to four years, achieved in just 12 months."

AI Adoption Lead — Wonga (former mentee)

What I do

Production AI systems with the boring bits done properly.

I focus on the parts of AI delivery that determine whether a system is actually useful after launch: trustworthy retrieval, observability, evaluation, release discipline, clear interfaces, and privacy-conscious architecture.

That means GenAI and RAG systems, MLOps and AI platform engineering, and delivery practices that survive contact with real users, real constraints, and real governance requirements.

GenAI & RAG engineering

Retrieval pipelines, grounded responses, tool orchestration, evaluation loops, and secure prompt workflows for production use.

MLOps & AI platforms

Automated training, deployment, monitoring, reproducibility, and scalable delivery patterns for model-heavy organisations.

Cloud architecture

Practical, security-aware cloud systems across container platforms, CI/CD, infrastructure-as-code, and controlled data boundaries.

Commercially useful AI

Systems tied to throughput gains, reduced manual work, better governance, and commercial outcomes rather than novelty theatre.

Site map in plain English

Start here, then follow the path that matches your interest.

If you are here to assess my profile, you can move from About to Projects, Writing, CV, and finally Contact within a couple of clicks.

About

Career history, domain background, delivery style, and why production-grade AI is my lane.

View profile

Projects

Case studies spanning GenAI, RAG, model monitoring, cloud architecture, and AI platform engineering.

Explore projects

Writing

Writing on MLOps, production AI systems, applied mathematics, ethics, and engineering judgement.

Read writing

Lab

A showcase of Generative AI apps that I've built

Explore Lab

CV & contact

Need the concise version? Download my CV or get in touch directly about consulting or roles.

Download CV Contact

Featured work

Projects that show how I approach production AI delivery.

These cover AI platform engineering, model lifecycle automation, retrieval-led experiences, and applied systems thinking.

Production RAG Chatbot (Enterprise application)

Led architecture and delivery of a production-grade RAG chatbot for John Lewis Partnership's internal workforce — from first line of code to 90% adoption in one week. Drummie answers HR, …

Python, Google Cloud Run, Gemini API, RAG, LangChain, Vector Store, Google Chat API, GCP

Read case study

The Prediction Factory: Designing an ML Platform from First Principles

Defined and delivered JUMO's internal ML platform from first principles — a config-driven orchestration layer that scaled model training from ~1 model per week to 50+ models per day, automated …

Python, Kubernetes, FastAPI, MLflow, Airflow, AWS S3, DynamoDB, EMR on EKS, Great Expectations, Datadog, PagerDuty, Docker

Read case study

Production ML Monitoring: From Weeks to Minutes

Designed and built a real-time model monitoring system at JUMO that reduced data anomaly detection time from weeks to minutes. The system automated batch scoring with Airflow-orchestrated Kubernetes jobs, computed …

Python, Airflow, Kubernetes, Great Expectations, Datadog, PagerDuty, AWS EKS, S3

Read case study

View all projects

Signal from noise.
Architecture that ships.

£40M+

50+ / day

~90%

70%

What people I've worked with say