I design, build and run systems in production, from the business problem to day-to-day operations.
For 19+ years I've taken technical responsibility for systems that have to work every single day: architecture, backend, integrations, infrastructure, security and AI applied as part of the engineering.
How I work
It started around 2007, still a teenager, tweaking templates and breaking pages on blogs until I understood what happened behind every line. By 14 it was already an obsession. More than 19 years later, that same habit of taking things apart to understand them is still my method.
I've lived through nearly every wave of the web: blogs and SEO, servers, PHP and Laravel, modern front-end, Docker, the cloud and now applied AI. Each phase added a layer: structure, performance, operations and finally architecture, which is where the decisions actually carry weight.
Today I'm accountable for whole systems: I design the architecture, build the backend and the integrations, provision the infrastructure, ship it, monitor it and keep it running. I've delivered multi-tenant SaaS and ERP platforms, real-time systems for thousands of concurrent users, industrial telemetry wired to PLCs and AI systems in production.
A large share of serious work happens inside software that already exists. I know how to step into a legacy codebase, recover the business rules nobody documented and modernise it in stages, without stopping operations and without breaking what already works.
Application security is part of that work. I review authentication, authorization, tenant isolation and API exposure, and I run technical security assessments of applications and APIs, including controlled vulnerability validation, always on my own systems or with authorization and scope defined by the system owner.
With AI I follow the same method as with any other technology. I use AI as an engineering component and as a multiplier for my own work, always with review, tests and attention to cost, security and the real limits of the models.
The timeline
The phases of the journey, in order, with each era's technologies in plain sight.
2007 First steps on the web. Where it all began: build, break and fix.
HTML · CSS · WordPress · PHP
2010 SEO, traffic & infrastructure. Drive traffic and keep it all up.
indexing · performance · servers
2013 Web engineering. Serious web apps with real business rules.
PHP · Laravel · APIs · multi-tenant
2016 Python expands the technical base. Automation, data and system-to-system integration.
automation · data · integrations
2018 Front-end & mobile. Modern interfaces and store-published apps.
React · Vue · Kotlin · Swift · Flutter
2020 Cloud, DevOps & architecture. Infrastructure as code and accountability for operations.
AWS · Azure · GCP · Docker · Terraform
2023 → now AI inside the engineering. AI as a system component, with tests, cost and limits. First experiments around 2019.
RAG · agents · LLMs in production
The technologies from earlier phases are still in use in current projects.
I take on the whole cycle and stay accountable for the system once it's live.
Understand the business rules and what actually needs solving.
Set structure, boundaries and trade-offs, with maintenance cost in view.
Build backend, interfaces and automations, with tests and review.
Wire up APIs, data, third-party services and field equipment.
Provision cloud, containers and infrastructure as code, with hardening and least privilege.
Ship through automated pipelines, with predictable rollback.
Metrics, logs, alerts and incident investigation when something goes off track.
Maintain, modernise legacy and evolve without breaking what works.
Real-world cases presented through the problem, responsibility and delivery stage.
Some projects are presented anonymously due to contractual confidentiality.
Regulated sectors needed their own management platform without running a separate system per client. I delivered multi-tenant platforms with billing, APIs and third-party integrations.
Responsibility: Multi-tenant architecture, domain modelling and integrations.
Event and class broadcasts with thousands of concurrent viewers, where interaction could not lag. I delivered the platform with real-time chat, polls and assessments.
Responsibility: Real-time architecture, backend and ongoing production support.
Validating identities, assignments and assets across distributed teams, tracking field journeys and handling anomalies required context and traceability that scattered tools could not give. I built an integrated platform from the ground up (web operations centre, API and mobile apps) with on-device authentication, geolocation, operational checks, real-time events and incident response.
Responsibility: Full-cycle architecture and delivery across backend, web operations centre, mobile apps and operational integration.
Administrative and commercial processes relied on manual work spread across applications with no proper integration. I designed an agent architecture able to read interfaces, drive web and desktop systems and coordinate flows with rules, approvals and traceability.
Responsibility: Agent architecture, task orchestration and integration between systems.
Operations had no visibility into what field equipment was doing. I integrated sensors and PLCs over LoRaWAN, MQTT and Modbus and delivered real-time operations dashboards.
Responsibility: End-to-end integration, from field equipment to the operations dashboard.
Managing properties, contracts, billing, maintenance and documents was scattered across processes that never talked to each other. I defined the architecture and the scope of a modular, multi-tenant platform to centralise property operations, financial flows and ecosystem integrations.
Responsibility: Product architecture, domain modelling and scope definition.
Catalogue, stock, pricing, reputation, logistics and finance for marketplace operations were tracked in scattered controls. I designed an architecture of specialised agents, with responsibility per role, permissions, shared context, human approval, auditability and exception handling.
Responsibility: Multi-agent architecture, integration model for external APIs and action governance.
Digital lending with little friction, yet with validation, security and traceability at every step. I structured the flow for sign-up, identity verification, document upload, contracting and integrated payments.
Responsibility: Journey architecture, integrations and security controls.
Too many documents for manual search, in a context where a made-up answer was unacceptable. I built the ingestion, structuring and semantic search pipelines with embeddings.
Responsibility: Ingestion pipeline, indexing and answer-quality control.
Environments assembled by hand and failures found by users before the team. I standardised the infrastructure with Terraform and containers and put monitoring and error alerting into production.
Responsibility: Infrastructure as code, deployment pipeline and observability.
A portfolio that could answer questions about my work from what is published, without keeping anyone's conversation. gilmar.dev is static, in three languages, and Gil answers with RAG over a curated knowledge base and the site's own content. She can also switch the theme and accessibility settings or open the résumé, always at the visitor's request. It is the one case in this portfolio you can try right now.
Responsibility: Product, architecture, front-end, assistant backend, infrastructure and deployment.
I design and ship systems with RAG, agents, document ingestion, embeddings and vector databases, automations and real-time voice, wired into the systems that already exist. I treat AI like any other engineering component: model evaluation, tests, cost and latency control, explicit limits and failure modes, human review wherever a mistake is expensive, data security and observability in production. I also use AI in my own process to produce more, without giving up review, tests and technical responsibility for what goes live.
I pick the tools after the architecture. This is the base I use to sustain it in production.
Got your message. I'll reply soon to the email you provided. If it's urgent, ping me on WhatsApp.
Reply goes to
The conversation stays in your browser; the server keeps no history. Your message is sent to OpenAI to generate the reply, and Cloudflare confirms you're not a bot. If you ask to get in touch, what you write reaches Gilmar by email (Web3Forms). Nothing is sold or shared.
Read the full policyGil is an AI and may be wrong. ·
Just a moment. Confirming you're human.