AI agents, under control
Connect AI to operational systems through MCP services, explicit permissions and audit trails. Put human review where decisions need it.
Architecture, tool integration, access controls and evaluation.
AI. Data. Delivered.
We design and build secure AI, data and cloud systems for organisations where reliability and control matter.
Led by Leonardo Martins.
An architect who still writes the code.
Built for the real world
From trading infrastructure to public services, the challenge is the same: make complex systems work, safely, in production.
25+ yearsEngineering distributed systems
UK patentGB2541570, ML-driven optimisation
Cloud Excellence2022 award-winning serverless work
Solutions with substance
Start with a business problem. Leave with a system your team can operate and build on.
Knowledge you can find
Turn disconnected documents and data into useful answers. Combine domain-aware retrieval, hybrid search and reranking with evidence and measured quality.
RAG pipelines, semantic discovery and evaluation harnesses.
AI spend you can govern
Make model usage visible and accountable. Introduce a gateway for routing, rate limits, token budgets and attribution across multi-model workflows.
Gateway integration, budget controls and cost reporting.
Data that joins the dots
Build a governed view of your operation across systems. Trace a journey from ingestion to reporting, identify bottlenecks and preserve the context behind each metric.
Data platforms, ingestion, lineage and operational insight.
Platforms that hold up
Stabilise unreliable services and modernise the components that constrain delivery. Combine distributed-systems experience with reproducible infrastructure and hands-on engineering.
Reliability reviews, cloud migration and performance engineering.
Designed.
Built.
Operated.
Where do we start?
Choose the challenge closest to yours. See a practical way to begin the conversation.
Start with one workflow
Choose a bounded use case. Map the tools, permissions and human decisions, then build a working integration with an evaluation plan.
A useful first output: a tested path from request to action, with the controls visible.
Experience that travels
Selected work from Leonardo’s career. Different environments, with the same responsibility to make the architecture deliver.
Public services
Architecture and hands-on delivery for the BI, MI and AI platform at His Majesty’s Passport Office, including serverless services and a unified data layer.
Terminal Repository / Passport Data Services received the 2022 Computing Cloud Excellence Award.
Capital markets
Led the architecture of HSBC’s first eFX co-location estate across London, New York and Tokyo. Designed sub-millisecond market-data access with a 25× reduction in latency spikes.
Earlier work at Barclays Capital included messaging infrastructure across 40 applications and seven business units.
Privacy engineering
Built Pontus Vision’s privacy platform to extract and connect personal-data entities across structured and unstructured sources, including documents and PDFs.
Person, Object, Location and Event modelling makes the relationships between records explicit.
Explore the privacy workAI infrastructure
Built MCP wrappers over operational data services and designed governance for multi-model pipelines. Financial-research RAG combines hybrid OpenSearch retrieval, finance embeddings, reranking and RAGAS evaluation.
Engineering spans Python, TypeScript and Go, with infrastructure defined through CDK, Terraform and Ansible.
Privacy is part of the system
Our roots are in finding personal data, connecting its context and making it manageable.
The Pontus Vision platform brought extraction, relationship mapping and compliance reporting together. That experience informs how we approach AI access, data lineage and governance today.
Explore the open-source project
I work across architecture and code, from the decisions that shape a platform to the details that make it run.
My career began in messaging middleware and capital markets. Today I build cloud, data and AI infrastructure, with the same attention to performance, reliability and operational reality.
I stay hands-on in Python, TypeScript and Go. I also work with the people who fund, assure and operate the service, so the engineering remains connected to the outcome.
UK Patent GB2541570
Executive MBA, London Business School
Accenture FinTech Innovation Lab winner, 2015
Thinking from the work
Practical perspectives on the systems behind enterprise AI, published on LinkedIn.
Token economics and the MCP gateway
Where routing, budgets and cost attribution belong in the AI stack.
Find the post on LinkedInSemantic discovery
Help people and agents find authoritative data products across a fragmented estate.
Find the post on LinkedInWhy “archiveloper” matters
Architectural judgement grounded in the experience of building the thing.
Read the article on LinkedInLet’s make it work
Bring the problem, the constraints and the ambition. We’ll work out a practical way forward.