Data, Analytics & AI Consulting
27 years building this — for USAID, the US Department of State, UK FCDO, the European Commission, and private-sector clients — often in places with no reliable mobile network, let alone a data center. Cloud, AI, and cybersecurity, done by the person who answers when it breaks.
End-to-end consulting
across the data stack
Applied AI & Agent Engineering
I was building machine learning models back when they were just called models — forecasting, classification, and the unglamorous statistical work still running underneath half the things people now call AI. So when a client asks for an LLM and what they need is a tuned classical model at a tenth the cost that doesn't hallucinate, I say so. On the generative side: agent orchestration with LangGraph and MCP, RAG pipelines, and routing between local open-weight models and cloud APIs based on how sensitive the data is. Knowing which problems don't need the expensive answer is most of the value.
- Azure AI Foundry
- MCP
- LangGraph
- RAG
- Computer Vision
- Classical ML
AI Automation
Automation either saves people hours a week or quietly corrupts a process nobody's watching closely enough. I've shipped both, early on, and the difference isn't the tool — it's knowing which decisions a model should never make unsupervised, and putting a human checkpoint exactly there. Power Automate, Logic Apps, Event Grid, Service Bus, agent orchestration: pick whichever fits the estate. The judgment about where to stop automating is the part that's hard.
- Power Automate
- Zapier
- Make
- Azure Logic Apps
- Agent Orchestration
Cloud Architecture
I've built physical data centers in the United States, the United Kingdom, Bolivia, Thailand, Iraq, and Pakistan. I've kept field offices online in places with no fixed lines and no dependable mobile coverage — the Chapare in Bolivia, the tribal areas of Pakistan — running on satellite links and whatever power we could generate. Now it's Azure estates, AWS where the client already lives, and migrations off legacy stacks under live donor contracts where the acceptable downtime window is roughly zero. Backups and disaster recovery get the same attention as the interesting parts — that's what separates a system that survives its first bad day from one that doesn't.
- Microsoft Azure
- AWS
- Google Cloud
- Bicep/Terraform
- AKS
- CI/CD
Data & Analytics
Most of this experience is USAID: monitoring and evaluation platforms, performance dashboards, forecasting models, and field data collection systems for programmes running across dozens of countries, over most of two decades. More recently the European Commission's Cancer Mission dashboards, tracking 60+ funded projects across seven analytical dimensions — and food-security forecasting for COMESA across East and Southern Africa. The stack is Power BI, Microsoft Fabric, Data Factory, SQL Server, Azure SQL, Cosmos DB, PostgreSQL/PostGIS, plus vector stores (pgvector) and graph databases (Neo4j) where the questions call for them — structured, semi-structured, and unstructured, depending on what the data actually is rather than what the tooling prefers. A dashboard nobody trusts is worse than no dashboard, because now there's a meeting about it. Trust comes from getting the data model right before anyone sees a chart — and from being able to explain it to the person signing off on the programme.
- Power BI
- Microsoft Fabric
- Data Factory
- AWS Glue
- PostgreSQL/PostGIS
Microsoft Platform
Most people learn Power Platform from documentation. I learned it building the systems programmes actually run on, where a field office in Yemen or Somalia depends on it working and "it worked in the demo" doesn't survive contact with that. Copilot Studio and AI Builder are new. Knowing what a low-code platform should never be trusted with isn't.
- Power Platform
- Power Apps
- Dataverse
- Copilot Studio
- Microsoft 365
- ALM
Cybersecurity
Security isn't a line item at the end of a proposal. It's in how access policies get written, how AI systems that touch sensitive data are designed, and who picks up the phone when something goes wrong. I did this work in places where a breach wasn't a compliance problem — it was a safety problem for the people in the dataset. That permanently changes how you weigh it. DORA, NIS2, the EU AI Act are new names on an old instinct about what actually needs protecting, and from whom.
- Entra ID Governance
- Purview DLP
- Secure SDLC
- OWASP
- Incident Response
- GDPR
Geospatial Analytics
700+ maps and geospatial products delivered under a single USAID contract, back when this meant building the systems by hand rather than bolting a model onto imagery. I know what the data looks like on the ground because for a long stretch, the ground is where I was — the Chapare in Bolivia, Peshawar, Baghdad, Islamabad, Bangkok, and now Tarapoto in the Peruvian Amazon. That shows up in knowing which satellite-derived answer to distrust.
- ArcGIS Pro
- QGIS
- GRASS GIS
- PostGIS
- RDF/SPARQL
- Remote Sensing
Strategic Advisory
Most technology strategy comes from people who've mostly advised. When I tell a leadership team their roadmap is unrealistic, or that a vendor's pitch doesn't match what the system underneath can actually do, it's not a hedge — it's from having held the pager when someone else's optimistic roadmap collapsed. Roadmapping, architecture review, technical due diligence, vendor selection.
- Tech Strategy
- Roadmapping
- Architecture Review
- Due Diligence
Programme & Project Management
Multi-year programmes rarely fail because nobody understood the technology. They fail because risk, budget, and stakeholder expectations weren't tracked at the same pace as the build — or because a team spread across four time zones and three working languages quietly stopped meaning the same thing by "on track." I've run both sides on programmes across three continents, including a $96M USAID monitoring contract, which means the project plan and the architecture diagram come from one person instead of getting lost in translation between two.
- Programme Management
- Risk Management
- Resource Planning
- Stakeholder Engagement
Turning architecture
into outcomes
I'm not an agency. I'm the architect, the builder, and the person who picks up when something breaks at 11pm in a country office eight time zones away. No account manager in between, no ticket routing to whoever's on shift that week.
Four continents, some of them mid-conflict. Long enough to have made most of the mistakes already, on someone else's dime, and stopped making them. When I say a migration needs a rollback window, it isn't caution from a textbook — it's from watching one go wrong without one.
Most of that work has meant running teams that were nothing like each other: engineers in Islamabad, analysts in Frankfurt, programme staff in Washington and London, all on one delivery. Different languages, different working cultures, different assumptions about what "done" means. Getting a distributed team like that to actually ship is a separate skill from the architecture, and it's one I've had a long time to get wrong and then get right.
All around the world, I built the IT and GIS departments from nothing — hired them, trained them, and handed over systems that kept running long after I left. That's the part of the job I care most about. Programmes don't need a consultant they can't get rid of; they need a capability that stays behind. If your team can't run the thing without me in six months, I haven't finished.
Backed by 27+ years across 45+ countries and engagements for USAID, the US Department of State, UK FCDO, the European Commission, and RAND Europe, delivered directly and through Tetra Tech International Development, Management Systems International (MSI), Experience International, and DAI.
Through SuarezCloud, my personal cloud and AI platform, I build and operate the infrastructure behind this AI-governance approach — a multi-channel gateway routing tasks between local and cloud models based on sensitivity and cost.
Good systems start with clear thinking. I design before I build.
Theory is cheap. I focus on what ships and what works in production.
A system is only worth what the people funding it understand about it. I've had programmes survive review because the data story landed with a non-technical audience in the room, and I've watched good work die because it didn't. Getting from the architecture to the boardroom isn't a soft skill in this work — it's the difference between a project that continues and one that doesn't.
Deep expertise,
broad coverage
Applied AI & Agent Engineering
- LLM Integration / Agentic AI
- Azure OpenAI · OpenAI · Anthropic · Gemini
- LangGraph · MCP · Local Models (Ollama, Llama, DeepSeek)
- Azure AI Search / Knowledge-Discovery Pipelines
- pgvector / Neo4j Knowledge Graphs
- RDF / SPARQL / SKOS Ontologies
AI Automation
- Power Automate (Cloud & Desktop)
- Azure Logic Apps / Event Grid / Service Bus
- Agent Orchestration with Human-in-the-Loop Checkpoints
- Webhook & Event-Driven Integration
Cloud Architecture
- Microsoft Azure (App Service, Functions, Durable Functions, API Management, Key Vault)
- Amazon Web Services (EC2, Lambda, S3, IAM, CloudWatch, Lightsail)
- Google Cloud Platform
- Infrastructure as Code (Bicep, ARM, Terraform, CloudFormation)
- CI/CD (Azure DevOps, GitHub Actions)
Data & Analytics
- Power BI (Authoring, Dataflows, Semantic Models, Embedded, Paginated)
- Microsoft Fabric / Azure Synapse / Data Factory
- PostgreSQL / PostGIS
- Azure SQL / SQL Server / Cosmos DB
- AWS Glue
- Python (FastAPI, SQLAlchemy, pandas, scikit-learn, statsmodels)
Microsoft Platform
- Power Platform (Managed Environments)
- Power Apps (Canvas & Model-Driven) / Power Pages
- Dataverse
- Copilot Studio / AI Builder
- SharePoint Online / Microsoft Graph / Teams
- ALM in Azure DevOps
Cybersecurity
- GDPR / DORA / NIS2
- ISO 27001 / SOC 2
- Secure SDLC / OWASP
- RBAC & Identity Governance (Entra ID)
- Purview DLP
- Third-Party Risk & WAF Incident Response
- ITIL Change & Incident Management
Geospatial Analytics
- Esri ArcGIS Pro / Online / Enterprise
- QGIS / GRASS GIS
- GDAL/OGR
- PostGIS
- Remote Sensing & Satellite Imagery Analysis
Strategic Advisory
- Technology Roadmapping
- Architecture Review & Technical Due Diligence
- Vendor Selection & Evaluation
Programme & Project Management
- Multi-Year Programme & Stakeholder Management
- Risk, Resource & Budget Oversight
- Distributed Multicultural Team Leadership
- C# / ASP.NET Core (.NET Modernisation) / TypeScript / PowerShell
Past Engagements
What collaborators
have to say
"The kind of IT professional you want on your team when you need fast, functional solutions that just work. Marcelo delivers fit-for-purpose solutions that balance functionality, budget, and context — exactly what you need in complex international development environments."
"His digital innovation skills and his ability to communicate complex data products clearly and convincingly to a variety of audiences saved our program from the chopping block. I recommend Marcelo in the strongest possible terms."
"One of the most competent systems digital technologists I've met, period. Marcelo's ability to integrate diverse digital systems and develop new ones in fast-paced and changing environments is truly exceptional."
"He has built whole IT and GIS departments from scratch in multiple projects he has worked on, which is a feat not many can achieve."
"His ability to translate complex technical concepts into practical solutions is unmatched. Marcelo consistently delivers high-quality work that not only meets but exceeds expectations, even in the most challenging environments."
"Consistently impressed by his exceptional Power BI skills, responsiveness, and innovative approach. He has a remarkable ability to translate complex data into clear, actionable insights and develop dashboards that meet the needs of stakeholders."
"He is a reliable and flexible colleague, fully responding to the needs of clients and teams, and delivering high-quality outputs. He integrates feedback while managing expectations of clients and teams."
Ready to solve
a hard problem?
Whether you need a scoped architecture review, a full project delivery, or an ongoing advisory relationship — let's start with a conversation.
Need ongoing infrastructure, hosting, or a SaaS platform instead of project-based advisory? See SuarezCloud.