DevOps Services in Paris, France

Aerospace / Manufacturing / DevOps / Paris

A Paris-based aerospace tier-1 supplier (EUR 4.8B revenue, manufacturing engine components, landing gear systems, and avionics housings for Airbus, Boeing, Safran, and Rolls-Royce, with 12,000 employees across 8 manufacturing sites in France, UK, Mexico, and India, running 220 production applications) needed a DevOps transformation to support its digital manufacturing strategy. The company's technology infrastructure had been built for a traditional manufacturing company — stable, conservative, and slow to change — but the aerospace industry's shift toward digital twins, predictive maintenance, and connected manufacturing required infrastructure agility that the current approach couldn't deliver. Core challenges: (1) Manufacturing system rigidity — the company's manufacturing execution systems (MES) deployed across 8 factories, each running slightly different configurations adapted to local production processes. MES updates required: 6-month planning cycles, factory-by-factory deployment during plant shutdowns (typically August and December), and manual testing at each site. An MES bug affecting quality data collection discovered in March would not be fixed across all sites until December — 9 months of manual workarounds by factory operators. New digital manufacturing capabilities (IoT sensor integration, real-time quality analytics, digital work instructions) waiting years in the deployment queue because each required the same 6-month deployment process as a bug fix. (2) Multi-site inconsistency — each factory running different versions of the same applications due to staggered deployments. Factory A running MES v3.2, Factory B running v3.4, Factory C running v3.1 (a deployment that failed and was rolled back). Version inconsistency causing: different quality data formats between sites (complicating group-level quality analytics), different feature availability (some factories having capabilities others lacked), and different bug profiles (bugs fixed at one site still present at others). Customer audits (Airbus, Boeing conducting regular supplier audits) noting inconsistency — "your quality system operates differently at your French and Mexican plants." (3) AS9100 and NADCAP compliance — aerospace quality standards (AS9100D, NADCAP for special processes) requiring documented change management for any system affecting product quality. The company's change management process: paper-based change requests, committee approval meetings (monthly), and manual deployment verification. Each deployment generating 40-60 pages of compliance documentation, manually prepared by quality engineers. The quality team spending 35 percent of their time on change documentation rather than actual quality engineering. ANSSI requirements: the company classified as an OIV (Operateur d'Importance Vitale — critical infrastructure operator) by ANSSI, requiring specific cybersecurity measures for information systems, including documented security controls for deployment processes. (4) Connected manufacturing latency — the company's digital manufacturing strategy required: IoT sensor data from 4,200 machines across 8 factories flowing to central analytics, digital twin models receiving real-time production data, and predictive maintenance algorithms processing vibration and temperature data. But the current infrastructure was cloud-averse — all applications running on-premises in each factory's local server room, with no cloud connectivity. Factory IT teams managing 8 separate infrastructure environments independently. Deploying a new IoT analytics capability required: provisioning servers at each factory, configuring network connectivity, deploying applications manually, and managing 8 independent instances. Lead time: 4-6 months per capability per factory. (5) Security and air-gap requirements — certain defence-related manufacturing (the company producing components for military aircraft programmes) requiring ANSSI-compliant security, including: network segmentation between defence and commercial production systems, specific access controls, and security monitoring. Some systems operating in partially air-gapped environments — creating DevOps challenges around deployment without internet access, dependency management in offline environments, and security scanning without cloud connectivity.

We delivered a DevOps transformation over 16 weeks — building a multi-site deployment platform that maintained aerospace quality compliance while enabling manufacturing agility. (1) Multi-site deployment platform: consistent, automated factory deployments. Unified platform: a centralised deployment platform managing all 8 factory environments from a single control plane — infrastructure-as-code (Terraform) defining each factory's infrastructure with site-specific variables (network ranges, hardware specifications, local integrations) while maintaining architectural consistency. GitOps: ArgoCD managing Kubernetes clusters at each factory site — manifests stored in Git as the single source of truth, with changes automatically synchronised to all sites. A change merged to the main branch deploying to all 8 factories within hours rather than months. Site-specific configuration managed through Kustomize overlays — the base application identical across all sites, with site-specific overlays for local integrations (different PLC brands, different ERP connections). Kubernetes at the edge: lightweight Kubernetes clusters (k3s) deployed at each factory site — running on existing server hardware, not requiring cloud connectivity for operation. Factory applications running locally for performance (MES cannot tolerate cloud latency), with synchronisation to the central platform for management and monitoring. Staged rollout: deployments following a staged pattern — pilot factory first (the French plant with the most mature IT team), then European factories, then international factories. Each stage including automated testing and manual verification before proceeding. Rollback capability: any factory can revert to the previous version independently within 5 minutes. (2) Aerospace-compliant CI/CD: quality and speed together. Pipeline architecture: every application having a standardised CI/CD pipeline with stages reflecting aerospace quality requirements: code analysis (SonarQube), security scanning (Trivy, Snyk), unit and integration testing, AS9100 compliance verification (automated checks for documentation, traceability, and configuration management), staging deployment (to a reference factory environment that mirrors production), automated acceptance testing (including manufacturing process simulation), quality approval gate (automated for standard changes, manual for safety-critical changes), and staged production rollout. Compliance automation: the 40-60 pages of manual change documentation replaced by automated generation — every deployment producing: complete change description, test results, security scan results, approval records, deployment timeline, and rollback plan. Documentation generated in French (meeting ANSSI and customer audit requirements) and English (for international sites and customer visibility). NADCAP traceability: deployment pipelines for special process systems (heat treatment, surface treatment, NDT) including additional verification gates — ensuring that software changes to quality-critical systems followed NADCAP-compliant change management. (3) Hybrid cloud architecture: factory edge with cloud intelligence. Architecture: factory operations running on local Kubernetes clusters (edge) for low-latency manufacturing operations, with cloud infrastructure (OVHcloud — SecNumCloud-qualified for defence-related analytics, AWS eu-west-3 for commercial analytics) providing: centralised data aggregation, advanced analytics (digital twin, predictive maintenance), cross-factory benchmarking, and group-level reporting. IoT pipeline: factory sensor data (4,200 machines) collected locally through edge processing (filtering, aggregation, anomaly detection), with summarised data synchronised to cloud for cross-factory analytics. Real-time alerting handled at the edge (sub-second latency for machine anomalies), analytical processing in the cloud (minutes for trend analysis). Hybrid connectivity: site-to-cloud connectivity through encrypted VPN tunnels with automatic failover — factory operations continuing independently if cloud connectivity was interrupted (edge-first architecture ensuring factory autonomy). (4) Security and air-gap deployment: ANSSI-compliant DevOps. Network segmentation: defence-related production systems on physically separated networks with controlled data diodes for one-way data flow (production data out, no inbound connections). DevOps for air-gapped environments: offline deployment packages built in the CI/CD pipeline, verified through checksums, and transferred via secure media to air-gapped networks. Container registry mirroring: an offline container registry at each air-gapped site, updated through periodic secure synchronisation. Images built and scanned in the connected environment, then transferred to offline registries with signed attestations of security scan results. Vulnerability management: security scanning results from the connected pipeline producing remediation packages for air-gapped environments — security patches deployable to offline systems within 72 hours of availability (versus the previous 30-day manual process). ANSSI compliance: all deployment processes documented against ANSSI security requirements for OIV operators — including: access control, audit logging, encryption, and incident response procedures. (5) Observability: manufacturing-aware monitoring. Monitoring: Prometheus/Grafana stack deployed at each factory (edge monitoring) with federation to a central Grafana instance (cross-factory visibility). Manufacturing-specific metrics: MES uptime, quality data collection success rate, machine connectivity status, and production scheduling accuracy. Factory dashboards: each factory having a dedicated operations dashboard showing: application health, deployment status, infrastructure capacity, and manufacturing KPIs. Factory IT teams self-sufficient for day-to-day monitoring while central IT maintains cross-factory visibility. Alerting: manufacturing-context-aware alerting — a MES slowdown during active production shift generating immediate alerts, while the same slowdown during planned maintenance being suppressed. Alert escalation reflecting factory shift patterns (French factories versus Mexican factory time zones).

Results over 12 months. Deployment velocity: deployment frequency from 2 per year (plant shutdown windows) to average 4 per week across all factories. MES bug fix deployment from 9 months to 3 days — the March bug now fixable and deployed to all 8 sites within a week. New digital manufacturing capabilities: deployment lead time from 4-6 months per factory to 2-3 weeks for all factories simultaneously. Multi-site consistency: all 8 factories running identical application versions within 48 hours of each deployment (versus up to 12-month version spread previously). Customer audits: Airbus and Boeing audits noting "consistent, well-managed technology infrastructure across all manufacturing sites" — a marked improvement from previous inconsistency findings. Quality data formats standardised across all sites — enabling group-level quality analytics for the first time. Compliance: AS9100 change documentation generated automatically — quality engineer time on change documentation from 35 percent to 8 percent of total time, freeing 4,200 hours annually for actual quality engineering work. ANSSI compliance maintained and improved — automated security scanning reducing vulnerability remediation from 30 days to 72 hours for air-gapped environments. NADCAP audit finding zero deployment-related nonconformities (versus 3 in prior audit cycle). Digital manufacturing: IoT sensor analytics deployed across all 8 factories within 10 weeks (would have taken 3+ years under previous approach). Predictive maintenance models for 1,200 critical machines — detecting anomalies an average of 72 hours before failure, preventing 34 unplanned machine stoppages in Year 1 (estimated production impact of EUR 8.4M avoided). Digital twin integration: production data flowing to digital twin models with sub-minute latency — enabling engineering teams to validate design changes against real production data. Security: ANSSI compliance documentation automated — annual compliance review from 6 weeks to 1 week of effort. Zero security incidents related to deployment processes (versus 2 in prior year). Air-gap deployment time from 30 days to 72 hours — critical security patches reaching defence-related systems 10x faster. Cost: infrastructure consolidation from 8 independent factory IT environments to a managed platform — infrastructure management cost from EUR 4.8M to EUR 3.2M annually. Factory IT headcount redeployed from infrastructure management to digital manufacturing support. Cloud analytics infrastructure: EUR 380K annually (new cost, funded from infrastructure savings). Financial impact: unplanned stoppage prevention EUR 8.4M; quality engineer productivity EUR 1.2M; infrastructure savings EUR 1.6M; deployment velocity (faster time-to-capability) EUR 2.8M; security improvement EUR 480K. Total annual impact: EUR 14.5M. Engagement cost: EUR 720K. Annual platform management: EUR 195K. ROI: 20.1x first year.