0 — Surface
Backend Engineer
Distributed systems, IoT platforms, and the invisible infrastructure that makes things work.
1 — Interface
The contracts that connect systems. REST endpoints, NGSI-LD for digital twins, LoRaWAN device provisioning, OAuth2 flows — every interaction starts at the interface.
The Digital Twin Framework in production: a single gateway unifying heterogeneous urban data sources — digital twin entities, IoT sensors, mobility and environment data — behind one consistent API, feeding a Unity-based 3D visualization of the city.
/ngsi-ld/v1/entities NGSI-LD Query digital twin entities with NGSI-LD filtering
/api/v1/ingest/lorawan LoRaWAN ChirpStack uplink webhook — idempotent ingestion into NATS JetStream
/api/v1/sensors/:id/timeseries REST Time-series readings served from TimescaleDB continuous aggregates
/oauth2/token OIDC OIDC token endpoint — RS256-signed JWTs, validated via JWKS
/api/v1/mobility/flows REST Mobility data unified by the gateway for 3D visualization
/api/v1/subscriptions/:id NGSI-LD Remove a real-time notification subscription
Context-based entity API for digital twin interoperability
Long-range IoT protocol for low-power sensor networks
Token-based authentication with client credentials and PKCE
Lightweight pub/sub messaging for real-time sensor data
2 — Services
Where business logic lives. Domain-driven services, clean dependency boundaries, and architectural patterns that keep complexity in check.
Component-based TypeScript framework for building NGSI-LD digital twins, built as an in-house replacement for FIWARE/Orion. Structured as a layered monorepo, open source under the MIT license.
Laravel platform delivered and running in production at ULB.
Language-exchange platform matching learners for tandem practice.
Application submission platform handling candidate files end-to-end.
Dependency injection and inversion of control for testable, decoupled service layers. Constructor injection, interface-driven contracts.
BullMQ workers processing async jobs — sensor data ingestion, entity updates, notification dispatch. Decoupled producers and consumers.
Command-Query Responsibility Segregation separating write paths (entity mutations) from read paths (optimized queries, cached views).
Layered workspaces with shared types, independent deployability, and dependency boundaries enforced by workspace constraints.
3 — Data
From sensor frames and external APIs to queryable entities. Idempotent multi-protocol ingestion on NATS JetStream with a dead-letter queue, time-series in TimescaleDB hypertables — and a BullMQ-orchestrated pipeline turning collected data into NGSI-LD entities.
Scheduled Collectors fetch data from external APIs and IoT sources
Cron / CollectorsJobs dispatched across 4 priority queues for async processing
BullMQ / RedisHarvesters aggregate and transform collected data into new datasets
Harvesters / EventBusMetadata and historical data persisted, raw files on S3-compatible storage
PostgreSQL / OVH S3Entity last-state and active subscriptions cached for sub-ms reads
RedisNGSI-LD API exposes entities with subscription-driven real-time notifications
NGSI-LD / REST4 — Infrastructure
The physical and virtual substrate — and the pipeline that ships to it. Two substrates: a Proxmox VM running the SPARC backend under Docker, and an OKD/OpenShift cluster running everything else behind a full GitOps chain where merging to main is the deployment.
Sensor Platform for Academic Research and Community — an IoT platform in production across ULB campuses. Carried end-to-end solo: Rust backend, Nuxt 4 web application, and its deployment.
A merge on main is the only deployment trigger
In-cluster pipelines build and push the images
Syncs cluster state to the Git repository, continuously
Per-environment overlays, no templating drift
Secrets encrypted at rest in Git, decrypted only in-cluster
Hardened workloads: readOnlyRootFilesystem, non-root containers, OpenShift SCC-constrained pods.
Containerized microservices with multi-stage builds. Compose for local dev, orchestrated deploys for production.
Runs the SPARC backend stack under Docker: Rust API, NATS, TimescaleDB, ChirpStack, Traefik. Event-driven TLS cert rotation, LoRaWAN chain secured gateway-to-network-server.
Long-range IoT gateway network. Payload decoding, device provisioning, sensor fleet management.
5 — Core
I build the systems that run behind the screen — backend services, data pipelines, IoT infrastructure. I care about clean architecture, honest engineering, and making things that actually work under load.
Since 2023 I'm a backend engineer at STARI (ULB & FARI) in Brussels, in the orbit of FARI — the AI Institute for the Common Good. Distributed systems, digital twins, and IoT platforms in production. I write Rust when correctness and performance matter, TypeScript when velocity matters, and I'm always looking for the right tradeoff between the two.
I'd rather understand one system deeply than skim ten frameworks.
From schema design to deployment pipeline — I ship end-to-end.
No premature abstractions. Complexity is earned, not assumed.
With the FARI AI Academy, I delivered a one-day training on AI for the common good to BELSPO, the Belgian federal science policy office.