Full-Stack Software Engineer · Leiria, Portugal
Francisco Gaspar
Full-stack software engineer building backend services, data integrations and the products on top of them.
At xgeeks I build for clients in industrial technology, electric energy and automotive, from AI agents and full-stack platforms to backend services and integrations. Before that, at Deloitte, I automated integration testing for one of Canada’s largest banks. On the side I build Soon, LongTable and CarMan.
Experience
Software Engineer · xgeeks
06/2022 – Present- Industrial technology client: built AI agents that browse websites like real users to audit usability and accessibility, plus the FastAPI job workers, MLflow eval suite and Next.js dashboard around them.
- Electric energy client: built OMIP’s marketplace for power purchase agreements across the NestJS API and the Next.js marketplace and admin apps, including real-time negotiation chat, contract search and regulatory reports.
- Automotive client: built backend services and third-party integrations in TypeScript and NestJS, with PostgreSQL and MongoDB data models in TypeORM and Jest tests.
- Deployed services on Azure with Terraform, monitored with Datadog and Azure Application Insights.
- Set up client repos for AI coding agents (Claude Code, Copilot) with agent instructions and handover docs.
TypeScript · Python · NestJS · FastAPI · TypeORM · MikroORM · PostgreSQL · MongoDB · Redis · Next.js · Tailwind CSS · OpenAI API · Azure · Terraform · Docker · Jest
Programmer · Deloitte
12/2020 – 06/2022- Automated integration test suites for one of Canada’s largest banks, replacing manual runs that took several hours per suite.
- Covered end-to-end workflows across core banking and insurance systems, including the data exchanged between them.
Education
Bachelor’s Degree in Computer Engineering
Escola Superior de Tecnologia e Gestão, IPLeiria
09/2016 – 07/2020 · EQF level 6
Professional Course in Computer Management
Colégio Dinis de Melo, Leiria
09/2013 – 07/2016 · EQF level 4
Projects
Side project · Live · In development
LongTable: restaurant reviews with friends
Friends rate restaurants on the website or in the iOS and Android app, publicly or just for friends. Places are one shared catalogue, so the same restaurant is the same place for everyone: a searchable feed, a map and a page per place. An optional WhatsApp integration uses your own account to read a group’s messages and react to them, bringing the reviews already posted there in, visible only to the group’s members.
Next.js · Expo · TypeScript · Supabase · Clerk · Sentry · Umami
longtable.world ↗Side project · Live · In development
Soon: bill reminders from Gmail
Reads bill emails in Gmail, extracts the amount and debit date (from the PDF when needed), and puts each payment day on your calendar with a reminder the day before. Nothing reaches the calendar unreviewed: uncertain values are left blank, never guessed.
Next.js · TypeScript · Supabase · Gmail API · Web Push · Docker
soon-tracker.com ↗Side project · In development
CarMan: your car’s maintenance history
Keeps a car’s full maintenance history (services, invoices, fuel and reminders) and hands it over when you sell the car. For iOS and Android, launching first in Portugal.
Expo · React Native · TypeScript · SQLite · Drizzle
Client project · xgeeks · 2025 – 2026
OMIP Marketplace: trading power purchase agreements
Marketplace for OMIP, the Iberian energy market operator, where companies list, negotiate and manage power purchase agreements. Across the NestJS API and the Next.js marketplace and admin apps, I built real-time negotiation chat and support tickets over WebSockets, contract search and filters on Meilisearch, Excel exports with contract history, regulatory reports for REN and ERSE, and an outbox that keeps emails and notifications from being lost when Redis is down.
NestJS · MikroORM · PostgreSQL · Redis · Meilisearch · Socket.IO · Next.js · Tailwind CSS · Azure · Terraform
Client project · xgeeks · 2026
WebLens: AI usability and accessibility audits
Platform for an industrial technology client where AI agents browse a website like real users and report its usability and accessibility issues. I worked across all of it: the browser agents (authenticated access, screenshots, gap detection), the FastAPI API and its queue workers (recovery of abandoned jobs, retries, a gate that blocks actions on evidence still being deduplicated), an MLflow eval suite for agent quality, the Next.js dashboard (workspace creation and invitations) and the Azure infrastructure in Terraform.
Python · FastAPI · SQLAlchemy · PostgreSQL · Next.js · Browser Use · MLflow · Azure · Terraform · Playwright
Team hackathon · Sep 2025
Quack as a Service: factory safety monitoring
Cameras check that workers wear the protective equipment each room requires before letting them in, identified by QR code. YOLO detects falls, AWS Rekognition flags signs of stress, and AWS Bedrock turns the events into safety insights on a live dashboard.
Python · FastAPI · SQLAlchemy · YOLO · OpenCV · AWS Rekognition · AWS Bedrock
Article · Feb 2024
Introducing AutoGen: enhancing LLM apps through multi-agent conversations
xgeeks blog post on AutoGen and how multi-agent conversations support applications powered by large language models.
blog.xgeeks.com/ai-agents-101-with-autogen ↗