ROSPIN-SAT-1: CubeSat on-board computer (ESA Fly Your Satellite!)
Full-Stack Developer & DevOps Engineer, Romanian Space Initiative (ROSPIN), European Space Agency
Software, infrastructure and quality strategy for a CubeSat OBC under ESA's Fly Your Satellite! Design Booster: layered RTOS firmware with CCSDS telemetry, an embedded CI/CD lattice, hardware-in-the-loop benches and telemetry anomaly detection.
Problem
Flight software for a CubeSat must be deterministic, testable without the physical hardware most of the time, and delivered by a distributed student/engineer team with ground-segment-grade process discipline.
Solution
A modular firmware stack (bootloader, FreeRTOS/Zephyr, middleware, flight app) with CCSDS-compliant TM/TC, containerised ARM toolchains in Jenkins + GitLab CI, QEMU-based early validation, a self-service HIL reservation platform and unified observability across OBC and ground systems.
My contribution
- Designed the layered firmware architecture and tuned FreeRTOS/Zephyr scheduling, multi-bus I/O (I²C, SPI, UART, CAN) and packet-based CCSDS TM/TC with error detection, retries and configurable timeouts.
- Containerised ARM cross-compilation and built Jenkins + GitLab CI pipelines: static analysis (Cppcheck, Clang-Tidy), unit tests (Unity, Google Test), coverage (gcov/lcov), nightly builds, auto-deploy to remote benches via Ansible with rollback.
- Provisioned Debian/Fedora build farms, staging networks and VPN gateways with Terraform + Ansible; hardened with image scans (Trivy/Clair), key-only SSH, SELinux/AppArmor, nftables and VPN/VLAN segmentation.
- Delivered a self-service HIL platform: Flask/React reservation portal with SSO/RBAC and board-health metrics, QEMU fault injection, Jenkins + Docker Compose regression suites.
- Fluentd + Prometheus + Grafana observability with Alertmanager SLOs; pandas/scikit-learn anomaly detection on historical telemetry; Doxygen/Sphinx living docs and onboarding workshops.
Stack
C
C++
Python
Bash
FreeRTOS
Zephyr RTOS
CCSDS TM/TC
Docker
Kubernetes
k3s
Jenkins
GitLab CI/CD
Ansible
Terraform
AWS
Prometheus
Grafana
ELK Stack
Fluentd
QEMU
OpenOCD
GDB
CMake
pytest
Flask
React
OpenVPN / IPsec
nftables
SELinux / AppArmor
Linux
Networking
Chef
scikit-learn
pandasLinks
On-board computer software
- Layered, modular firmware (secure bootloader, RTOS kernel, middleware and flight application) tailored to CubeSat avionics.
- FreeRTOS/Zephyr integration for deterministic scheduling and inter-task communication over I²C, SPI, UART and CAN.
- CCSDS-compliant telemetry/telecommand services with error detection, retry logic and configurable timeouts for reliable up/downlink.
Embedded CI/CD and ground segment
- Dockerised ARM toolchain with Nexus/Artifactory artifact management.
- Jenkins and GitLab CI pipelines: Cppcheck and Clang-Tidy static analysis, Unity/Google Test unit tests, gcov/lcov coverage, nightly builds and merge-request validation; Ansible deployment to lab benches with rollback and Slack/Teams + Grafana feedback.
- GitOps guardrails: protected branches, mandatory peer-reviewed merge requests, pre-commit clang-format, semantic-version tagging; pytest + CMake smoke, regression and fuzz suites for the telecommand parsers.
Infrastructure and security
Terraform + Ansible for Debian/Fedora build farms, staging environments and VPN gateways; Trivy/Clair image scanning, SSH key-only auth with port knocking, SELinux/AppArmor confinement, nftables, and OpenVPN/IPsec with VLAN segmentation isolating development, test and payload networks.
Hardware-in-the-loop bench
As Full-Stack Developer, a reservation platform (Flask backend, React frontend) with SSO, role-based access and live status of multiple embedded boards; QEMU hardware emulation with fault injection for validation before hardware access; HIL regression suites orchestrated with Jenkins + Docker Compose covering sensor simulation and mock telemetry.
Observability and anomaly detection
Fluentd logging and Prometheus metrics across OBC and ground-station subsystems, Grafana dashboards, Alertmanager SLA/SLO rules; Python (pandas, scikit-learn) models trained on historical telemetry to flag subsystem drift early. Living documentation in Doxygen and Sphinx/ReadTheDocs with Conventional-Commit changelogs.
