Full-stack engineer with 8+ years across backend, frontend and applied AI, in cybersecurity, financial services, retail analytics, energy and telecom. At IBM I build Python services, retrieval-augmented generation features and the front-end tooling around them for QRadar EDR, an endpoint security product. I own solutions end to end, from data model and API through to the screen people use.
Applied AI, backend services and internal tooling for QRadar EDR, an endpoint security product.
Show the detailsHide the details
QRadar EDR (formerly ReaQta) is an endpoint detection and response product with MITRE ATT&CK-mapped detection rules, deployed on OpenShift and Kubernetes in cloud and air-gapped environments.
Built a retrieval-augmented generation platform over the product documentation and detection-rule corpus, so engineers can ask questions in plain language and get answers traceable to their source: document ingestion and chunking, embeddings indexed in OpenSearch, and hybrid semantic and keyword search with reranking.
Orchestrated multi-step, LLM-powered pipelines over that corpus using structured prompt templates, a thinking, drafting and checking workflow, and validated structured outputs. These power code generation, code explanation, similarity search against existing rules, and duplicate-rule detection.
Integrated a quantized open-weight model (GGUF via llama-cpp-python) into the product, bringing these capabilities to air-gapped customer environments on CPU-only infrastructure.
Built an interactive model evaluation platform where engineers load models, tune inference parameters (temperature, top-k, top-p, context window) and compare outputs in real time to support model selection.
Designed and delivered a validation service for Lua detection rules in three iterations: a FastAPI proof of concept orchestrating Luacheck and lua-format, integration and containerization within an existing service, then a production microservice with asynchronous subprocess handling, fail-fast error handling, type hints and full unit test suites. Detection engineers now use it to catch rule errors before release.
Built a smoke-testing service with a sandboxed Lua runtime, fully API-compatible with the endpoint agent, so detection engineers can validate policies before production deployment.
Added features to and maintained Yogi, the platform's API gateway, in Node.js and TypeScript, and built React and TypeScript front-ends for internal tooling, owning features from the data model through the API to the on-screen experience.
Maintain and extend the platform's microservices: harden container images to security standards, run performance profiling and stress tests, and manage the CI/CD pipelines, first on Jenkins and later on SPS.
Provide support on incoming cases, triaging by severity and impact and implementing fixes in priority order.
NetCapital generates leads for banks, credit card providers and other financial institutions across multiple European countries.
Gathered requirements directly from stakeholders for the rebuild of the company's internal CMS, translated them into technical specifications and distributed tasks across the team.
Built the back-end of the new CMS in Node.js and TypeScript, and the React front-end, later introducing Next.js. Structured the front-end to support multiple site designs, including Webflow-based designs integrated via DevLink.
Defined and executed the data migration from DynamoDB to MongoDB with preserved integrity, and documented the strategy for future reference and audit.
Introduced a caching layer to reduce page load times, and integrated Google Tag Manager so tracking tags could be managed without code changes.
Set up the development pipeline with GitLab CI and Docker for consistent deployments across environments, and investigated and resolved production anomalies.
Built a B2B e-commerce platform for a wholesale distributor, with a Django REST Framework back-end and a React and Redux front-end covering catalogue, stock, ordering and back-office order and complaint management, deployed on Heroku.
Developed a Python desktop application (Pandas, Tabula, PyInstaller) for an accounting firm that extracts, cleans and reconciles client transaction data and exports it into the firm's accounting software (Meteor Compta).
Worked directly with the accountants to gather requirements, delivered the solution first for the firm's main client, then generalised it to support additional clients.
Devoteam started a proof of concept to automatically match incoming CVs to missions. My team owned feasibility: choosing and validating the processing stack.
Set up a Hadoop cluster with YARN as resource manager, and tested job submission.
Installed Spark on the cluster and validated batch and streaming jobs.
Set up Apache Livy for interactive and batch jobs submitted from remote hosts.
Building scrapers for a retail market intelligence platform.
Show the detailsHide the details
Datagram runs a retail market intelligence platform serving clients such as Michelin and Carrefour.
Developed and maintained site-specific web scrapers (Scrapy, Selenium with headless Chrome, Beautiful Soup) running on a distributed worker architecture, including handling of rate limiting and anti-bot measures.
Routed collected data to Cassandra via RabbitMQ and loaded historical data into PostgreSQL, with integrity and data-type checks.
Built and scheduled batch workflows with Apache Airflow, and automated scraper operations with shell scripts.
Maintained and extended the Kubernetes-based microservices back-end serving the platform's dashboards.
Prediction pipeline automation, then design and development of an API.
Show the detailsHide the details
COR-E provides decision support for the wholesale electricity market: price forecasting and explanation of the key drivers behind each prediction.
Built a Flask REST API (MVC, SQLAlchemy, Swagger) exposing production data to the dashboard, with Redis caching of the latest predictions to reduce database load.
Automated model training and prediction pipelines triggered by weather-data availability, using shell scripts, cron and Pentaho Data Integration.
Introduced multiprocessing and multithreading into model and prediction scripts in collaboration with the machine learning team.
Designed the target database and maintained the SQL views feeding the forecasting dashboard.
Python, Flask, SQLAlchemy, Redis, PostgreSQL, MySQL, Pentaho DI
01/2015 – 12/2017
TOPNET
Technical support and reporting solutions.
Show the detailsHide the details
TOPNET is a major Tunisian internet service provider.
Provided technical support to residential and business customers, diagnosing connectivity, configuration and service issues through to resolution.
Built automated KPI reporting solutions for the technical customer support and marketing departments, tracking service quality and supporting data-driven decisions, and documented both solutions.
Python, MySQL, Power BI
Beyond the day job
Projects I built and articles I wrote on my own time.
A bilingual comparison site for Dutch financial products: credit cards, insurance, bank accounts, savings and investments. Server-rendered for search, with product pages that take on the look of the product they describe.
A self-hosted AI workbench that answers questions and writes code from a private knowledge base. Retrieval-augmented generation end to end, with every model running on local hardware.
Managing Database Migrations for Multiple Services in a Monorepo with Alembic
Article on dev.to
How to run database migrations for several services in one monorepo with Alembic: a separate migration history per service, selected through dynamic configuration, so each service migrates in isolation without touching the others.
Skills
The tools I reach for, grouped by where they sit in a system.
Languages
Python
TypeScript
JavaScript
Lua
SQL
Bash
Scala
Backend
FastAPI
Flask
Django
Node.js
Frontend
React
Next.js
Redux
Tailwind CSS
Data
Spark
Hadoop
pandas
Airflow
Kafka
RabbitMQ
Scrapy
Selenium
Infrastructure
Docker
Podman
Kubernetes
OpenShift
Terraform
AWS
Google Cloud
IBM Cloud
Heroku
Jenkins
GitLab CI
Artifactory
Git
Databases
PostgreSQL
MongoDB
MySQL
Redis
OpenSearch
Cassandra
DynamoDB
Turso
AI and LLMs
RAG
Context engineering
Agentic workflows
n8n
MCP
Embeddings and vector search
LLM integration
Hugging Face Transformers
llama-cpp-python
GGUF quantization
Inference tuning
Model evaluation
Air-gapped deployment
Education
2018
ESPRIT School of Engineering
Engineering Degree in Computer Science (BI and Data Science), EUR-ACE and CDIO accredited
2013
Institut supérieur de l'informatique
Bachelor's degree in network administration and services
Badges
IBM Build Smarter AI with Embeddings
IBM, 2026
Building Trustworthy AI Enterprise Solutions
IBM, 2026
IBM Developer Profession
IBM, 2026
IBM Big Data Developer Mastery Award
IBM, 2017
Contact
Hiring for a full-stack or AI engineering role, or have a project in mind? Write to me here or reach me directly.