08/25/2026 updated
TC
100 % available
Software Engineer | Agentic Development & Back-End Specialist
Torres Novas, Portugal
Only remote
BSc Software Engineering (Expected 06/2027) - IP de Castelo BrancoAbout me
Software engineer who multiplies his output with agents: Claude Code as daily driver, custom MCP servers so it works on the real codebase. Strongest on the back end — Python services, REST APIs, pipelines. I run my own web services business, and I maintain what I ship rather than hand it over.
AI AgentsRAGHTMLJava (Programming Language)JavaScriptHTML5Cascading Style Sheets (CSS)Code GenerationCode ReviewLinuxDjangoGitHubPython (Programming Language)Node.jsSQL Databases
I am a software engineer who builds with agents, and the two are not separable in how I work: engineering judgement decides what gets built and whether it is correct, the agent decides how fast, never the reverse.
AGENTIC DEVELOPMENT. Claude Code is my development environment rather than an assistant I consult. I work spec-first: the change is scoped in writing, planned, then executed in diffs small enough that review is real rather than nominal. Most of the leverage is in context engineering, not prompting — repos structured with CLAUDE.md conventions, custom slash commands for recurring operations, subagents for work that should not pollute the main context, and hooks that enforce guardrails deterministically instead of asking the model to remember them. I build custom MCP servers so the agent reads real state: filesystems, internal APIs, databases. That is the line between code generation and a working system — the model stops guessing about the codebase and starts operating on it.
LLM ENGINEERING. I build on the Anthropic API: tool use, structured JSON output that downstream code can depend on, retrieval over local datasets. At Código Luminoso this became internal automation tooling around LLM workflows, replacing recurring manual operations with scripted, reviewable processes — model steps inside a pipeline with defined inputs, outputs and failure behaviour, not a chat window.
VERIFICATION. I treat model output as a claim to be tested, not an answer: prompt iteration against concrete cases, regression checks, documented failure modes. At IBM/Softinsa I used LLM-assisted workflows to extend test coverage and reason through legacy code I had not written, with every change human-reviewed before merge. An agentic workflow without a review loop is only a faster way to ship defects.
THE ENGINEERING UNDERNEATH. None of the above holds without being able to read and judge the code. Python is my core language (Flask, FastAPI, Django), with SQL/MySQL, Pandas, Docker, Git, Linux and unit testing around it, and JavaScript/React on the front end. At IBM/Softinsa I built a data-filtration service for SAP in Flask, published as a documented REST API for downstream teams — the contract was the deliverable — over Pandas pipelines designed on the assumption that enterprise exports arrive malformed and schemas drift, so validation and explicit failure paths were part of the design rather than a response to the first incident. I worked to the team's existing code review and branching conventions.
OWNERSHIP. Through terraroot I run a web services business end to end: scoping with non-technical clients, build, hosting, deployment, then maintenance under retainer. Owning software for years rather than until handover taught me to scope conservatively, document decisions and stop when the foundation is wrong: I discontinued a SaaS prototype once I established that the third-party API beneath it could not meet a core requirement.
Experience
Professional experience0 - 5 years
Freelancer since2026
Languages
EnglishFluentPortugueseNative speaker
Project history
Founded and runs a web services business, building, hosting and maintaining client websites end to end from first scoping call through deployment to ongoing monthly maintenance. Owns the full delivery lifecycle under recurring retainer agreements including hand-built sites, hosting, domains, security updates, content changes and quarterly reporting. Ships production sites meeting real business requirements with mobile-first responsive layouts, SEO structure, GDPR-compliant contact handling and accessibility basics.
Built and shipped production websites end to end, both the company's own site and client sites delivered as a paid service. Built internal automation tooling around AI/LLM workflows, replacing repetitive manual steps in day-to-day operations with reviewable, scripted processes. Designed and implemented internal REST APIs so internal tools and systems could exchange data instead of relying on manual handoffs.
Built a data-filtration service for SAP in Python (Flask), exposing a documented REST API for downstream consumption by other teams. Engineered CSV processing pipelines with Pandas to transform and validate enterprise datasets. Delivered production-oriented back-end code inside an enterprise engineering team, working to their code review and branching conventions. Used LLM-assisted workflows to extend test coverage and reason through unfamiliar legacy code.