CV

Machine learning & AI solutions engineer — London, UK

Summary

I help customers work out what’s actually possible, and what’s actually feasible, then build a small proof of concept that shows them it works. That’s the part of the job I like most, and it’s taken me somewhere new every few months for close to a decade.

The environments are usually the awkward kind. Kubernetes onto a submarine, air-gapped because there’s no bandwidth underwater. A border control platform I supported from a secure room. Gene therapy pipelines demoed straight back to the customer that cut projected analysis costs by 80%. Lately it’s machine learning: a multi-agent platform for drug discovery, and fine-tuning Google’s MedGemma to generate radiology reports from breast scans it had never seen.

I know these systems well enough to build the thing under pressure and explain it at any level, from the code up to the business case. SC cleared, 12 AWS certifications, London.

Experience

Machine Learning Consultant — Independent

July 2025 – present

ML engineering and advisory for early-stage startups: multi-agent systems, LLM tooling, and MLOps.

Machine Learning Engineer — BioX AI

July 2025 – December 2025 · London

Multi-agent drug discovery platform. Deployed multi-agent workflows on Vertex AI Agent Engine using Google ADK, with A2A for inter-agent communication.
Agent tooling. Built agents for literature review, patent drafting and experiment planning; integrated MCP tools for real-time literature and patent search.
Vertex AI training. Fine-tuned Google’s MedGemma on DICOM breast scans to generate radiology reports, evaluated against BI-RADS scores as ground truth.
MLOps & infrastructure. Provisioned reproducible GCP environments with Terraform, set up cost tracking, and monitored agent runs with Agent Engine tracing and logs.

Machine Learning Engineer — The Machine Learning Institute

May 2025 – July 2025 · London

Accepted into the ML Institute AI Residency — chosen from 800+ applicants for a fellowship fully funded by the UK government, focused on moving applied deep learning from research into production.

RLHF with PPO. End-to-end RLHF pipeline for summarisation: reward model training, PPO from scratch (token-level KL divergence, reward shaping, advantage estimation), and parameter-efficient fine-tuning with LoRA/adapters.
Whisper emotion detection. Fine-tuned OpenAI Whisper to classify emotions and environmental sounds, using token-based training that mixed pre-training and new data to mitigate catastrophic forgetting.
Vision transformer captioning. Built a synthetic dataset and trained a ViT + transformer decoder for image captioning with CLIP-style loss.
Dual-encoder retrieval. Trained dual- and cross-encoder architectures with contrastive loss for semantic document retrieval; deployed with a Redis vector DB for ANN search.
Foundational work. Implemented Word2Vec and custom transformers from scratch.

Senior Consultant, Solutions Architect — HeleCloud (a SoftwareOne company)

April 2022 – June 2024

Deployed a Terraform-based AWS Landing Zone into customer environments, provisioning a full multi-account organisation from a single root account — with an account vending machine for ephemeral developer environments, centralised logging and billing, and isolation that contained the blast radius of any one project.
Led 10+ AWS Well-Architected Reviews end to end, running multi-day workshops with up to 12 stakeholders across security, operations, infrastructure and finance, then translating findings into prioritised remediation plans that cut high-risk issues by 45% — the threshold that unlocked AWS funding for the remediation work.
Migration specialist for 10+ AWS Optimisation and Licensing Assessments. Deployed data collection tooling across mixed on-premise estates (Hyper-V, vSphere, VMware), connecting at hypervisor or VM level over WMI, SNMP or T-SQL depending on customer access constraints, with workload data encrypted locally before syncing to S3.
Analysed 30 days of collected workload data to produce rightsizing and licensing migration recommendations, reducing TCO by 36%.
Built cross-account AWS Security Hub aggregation with filtered critical alerts and Microsoft Teams integration.

Senior Consultant, Platform Engineer — Automation Logic

May 2021 – April 2022 · London

Built a digital twin of an air-gapped environment in AWS with Terraform and CI/CD in ~10 minutes, and developed an offline Kubernetes deployment solution with Ansible + Nexus for air-gapped defence infrastructure.
Provided production support, on-site in a secure room, during deployment of a 4,000+ node Kubernetes platform (Europe’s largest at the time) — monitoring pods to ensure zero downtime for an automated border control system.
Built a workforce analytics dashboard on GCP (BigQuery + Data Studio) to analyse consultant–project alignment and reduce turnover.

Consultant, DevOps Engineer — PA Consulting

December 2019 – May 2021

Built a real-time data ingest pipeline with AWS Kinesis feeding an S3 data lake; implemented fine-grained access controls and exposed curated datasets to downstream partners via API Gateway + Lambda.
Integrated the S3 data lake with SageMaker so AI/ML teams could run experiments on gene therapy data, cutting projected analysis costs by 80%.
Created the Git repo and Terraform CI/CD pipelines provisioning and deploying into isolated dev/test/prod environments.

Cloud Engineer, AWS re/Start — Amazon Web Services

July 2019 – October 2019 · London

Full-time cloud engineering training: hands-on work with Lambda, DynamoDB and API Gateway alongside project management and customer service. The capstone project was a serverless website — www.roshbeed.com, the ancestor of this one — built on CodeBuild, CodeCommit and S3, with build time optimised to 30 seconds. Invited back after graduating as a guest speaker and mentor for new participants.

Consultant, Software Engineer — Zaizi

August 2017 – February 2018 · London

Built a geo-replicated distributed storage platform on AWS (EC2 + EBS + GlusterFS), giving 3,600 active users scalable, fault-tolerant, cross-region access to 4.6 TB of unstructured data (14M documents).
Replaced a manual Excel + PowerShell workflow with a Python pipeline that extracted, transformed and validated CSV data, cutting runtime from two days to two hours (~£100k annual saving).

Lead Software Engineer — Heka

May 2017 – August 2017 · London

Led a five-engineer team building a sports marketplace (Java microservices + AngularJS).
Migrated legacy infrastructure from DigitalOcean to GCP App Engine, cutting build time from six hours to ten minutes.

Software Developer — East City Films

February 2017 – April 2017 · London

Built a VR 360° video player with in-headset menu navigation.

Education

The Machine Learning Institute	Level 5 Digital Skills Bootcamp, Machine Learning
Amazon Web Services	Level 3 Digital Skills Bootcamp, Cloud Computing
University of Plymouth	BSc, Computer Science
Corndel	Level 5 Apprenticeship, Leadership and Management
Generation UK & Ireland	Cloud Support Program

Certifications & clearance

12 AWS certifications, including AWS Certified Solutions Architect – Associate, and AWS Partner: Accreditation (Technical). SC cleared.