03 / experience
Work Experience
My professional journey and key contributions
01 / role
Backend Developer
Legacy System Migration (AS400 → Laravel Microservices)
Active participation in the progressive migration of the main system based on AS400 and DB2 to a microservices architecture using Laravel, Vue, and Tailwind.
Delivered 500+ features across 20+ microservices by applying DDD principles and Ports & Adapters architecture, contributing to the full migration of 100+ AS400 modules off the legacy system.
Enabled reliable asynchronous communication between 20+ microservices by integrating RabbitMQ into a shared internal library, applying choreography, command/fact events and SAGA patterns for distributed error handling.
Reduced duplicated code across 20+ microservices by building a shared library with global Value Objects and a query DSL with filtering, improving query performance by 42%.
Cut static analysis issues by 62% and critical SonarQube issues to 0 by introducing SonarQube and PHPStan level 4 into the development workflow.
Reduced deployment time from 6+ hours to 2 hours by automating CI/CD pipelines with Jenkins and containerizing microservices with Docker.
Shortened incident diagnosis time by centralizing logs with the ELK stack (Elasticsearch, Logstash, Kibana).
Saved half a week per cycle and reduced operational errors by automating a previously manual process.
Price Discrepancy Detection System
Internal tool built with FastAPI and Vue.js that detects and analyzes discrepancies between internal and supplier prices.
Replaced 20 hours per month of manual price review by building an automated dataset comparison service with FastAPI, processing 1000+ prices per run.
Identified 450+ anomalies and €10,000+ in price inconsistencies by generating structured reports with the Gemini language model, improving traceability of commercial discrepancies.
Batches Migration & Architecture Improvement
Migration of batch processes from AS400 to a modern Laravel environment using Inertia, React (TypeScript) and Redis.
Migrated 200+ batch processes from AS400, redesigning business workflows to reduce legacy dependency and improve maintainability.
Processed 100+ jobs per day with a 99%+ success rate by implementing an asynchronous queue system with Redis, Laravel jobs and the scheduler.
Reduced time to detect failed emails and scheduled tasks from 3 hours to 15 minutes by building automated monitors and error dashboards with shadcn/ui.
Internal Ticket & Projects Analytics Platform
Analytics platform giving IT project managers a global view of tickets and projects through KPIs and interactive dashboards, built with FastAPI and React (TypeScript).
Delivered the full-stack platform in 3 days by orchestrating a multi-agent AI workflow with Claude, using 5 specialised roles (tech lead, backend, frontend, data visualisation and code quality), each with its own documentation and defined communication protocol.
Minimised architectural rework during implementation by running structured AI-assisted design sessions to define the architecture, data model, design system and patterns before writing code.
Achieved a maintainable, scalable codebase with 90%+ test coverage and 0 Ruff/ESLint errors by applying clean architecture, SOLID principles and an automated code-quality review agent.
Gave project managers visibility over 10,000+ tickets and 70 projects across 5 IT teams by building interactive dashboards with KPIs, filters and trend analysis.
side quests
Personal Projects
Things I've built exploring ideas on my own time
A web tool that analyses source code to estimate its time and space complexity (Big-O) and explains the results with AI.
- Estimated time and space complexity of Python code by building a static analysis engine based on AST traversal, with dedicated visitors for loops, recursion and allocation.
- Reduced response time for repeated analyses from 100 ms to under 2 ms by implementing a Redis caching layer as a reusable mixin across services.
- Made complexity results understandable for learners by integrating Groq-hosted LLMs to generate step-by-step explanations of each Big-O estimate.
- Built secure user accounts with JWT authentication and bcrypt password hashing, on a feature-based FastAPI architecture with MongoDB.
Wine Classifier — KNN from Scratch
A command-line K-Nearest Neighbors classifier built from scratch with NumPy to classify wines by type from their chemical features, as a hands-on way to learn the algorithm.
- Implemented KNN from scratch using only NumPy — Euclidean distance, k-nearest lookup via argsort, and majority-vote classification with a closest-neighbor tiebreaker.
- Applied min-max feature scaling so no single feature dominates the distance calculation due to a larger numeric range.
- Validated the model with leave-one-out cross-validation across odd k values from 1 to 15, reporting per-k accuracy to find the best fit.
- Built an interactive CLI with a classification mode and an --evaluate mode for comparing k values on the wine dataset.