IoT-Enabled Freight Tracking & AI Logistics Platform
Challenge
A Swiss IoT-first freight operator needed to upgrade its cargo-tracking stack to keep pace with global competition. The legacy monolith could ingest device pings from railcars and containers, but dashboards refreshed only every few hours, alert rules were hard-coded, and any UI tweak risked firmware regressions. Customers demanded minute-level ETA updates, AI-powered anomaly alerts and a modern self-service portal. At the same time, the client’s fleet of LTE/5G trackers was growing fast, pushing daily message volume beyond ten million. The new solution had to roll out in under a year, run both on-prem and in the public cloud for data-sovereignty reasons, and integrate seamlessly with an existing order-management ERP
A Swiss IoT-first freight operator needed to upgrade its cargo-tracking stack to keep pace with global competition. The legacy monolith could ingest device pings from railcars and containers, but dashboards refreshed only every few hours, alert rules were hard-coded, and any UI tweak risked firmware regressions. Customers demanded minute-level ETA updates, AI-powered anomaly alerts and a modern self-service portal. At the same time, the client’s fleet of LTE/5G trackers was growing fast, pushing daily message volume beyond ten million. The new solution had to roll out in under a year, run both on-prem and in the public cloud for data-sovereignty reasons, and integrate seamlessly with an existing order-management ERP
Solution
Sigli embedded a cross-functional squad and delivered the platform in four waves:
1. Architecture sprint. We decomposed the monolith into domain-driven micro-services, drafted SLAs and chose Kotlin + Spring Boot for low-latency ingestion plus Angular 16 + Nx for a modular SPA.
2. IoT data plane. MQTT and gRPC gateways now stream tracker telemetry into a Kafka backbone; time-series data lands in TimescaleDB, while business events surface via WebSockets to the portal.
3. AI communication layer. A lightweight generative-AI micro-service (OpenAI-compatible LLM fine-tuned on logistics chat logs) converts raw alerts into plain-language recommendations for planners, exposed through a React-style chat widget.
4. DevEx & rollout. GitHub Actions → Helm charts automate blue-green K8s deployments to either EKS or on-prem Rancher; Terraform modules standardise VPCs, secrets and monitoring. Paired coding sessions and run-books ensured the in-house team could own the stack post-launch.
Sigli embedded a cross-functional squad and delivered the platform in four waves:
1. Architecture sprint. We decomposed the monolith into domain-driven micro-services, drafted SLAs and chose Kotlin + Spring Boot for low-latency ingestion plus Angular 16 + Nx for a modular SPA.
2. IoT data plane. MQTT and gRPC gateways now stream tracker telemetry into a Kafka backbone; time-series data lands in TimescaleDB, while business events surface via WebSockets to the portal.
3. AI communication layer. A lightweight generative-AI micro-service (OpenAI-compatible LLM fine-tuned on logistics chat logs) converts raw alerts into plain-language recommendations for planners, exposed through a React-style chat widget.
4. DevEx & rollout. GitHub Actions → Helm charts automate blue-green K8s deployments to either EKS or on-prem Rancher; Terraform modules standardise VPCs, secrets and monitoring. Paired coding sessions and run-books ensured the in-house team could own the stack post-launch.
Results
Real-time tracking ≤ 15 s lag for 9 400 railcars & containers (previously 10 min+) — thanks to the Kotlin event-stream pipeline.
AI alerting slashed incident triage time by 45 %; planners receive natural-language summaries instead of raw sensor codes.
User satisfaction ↑ 30 pp: revamped Angular dashboard and chat-style assistant replaced legacy SOAP portal.
Release cadence doubled — CI/CD shrank average time-to-feature from monthly drops to bi-weekly sprints.
Cloud-agnostic & future-proof: micro-services run identically on Azure AKS or on-prem Rancher, meeting Swiss data-residency rules while keeping a path open to full public-cloud scale. Together these gains position the client as an innovation leader in European smart-freight logistics.
Real-time tracking ≤ 15 s lag for 9 400 railcars & containers (previously 10 min+) — thanks to the Kotlin event-stream pipeline.
AI alerting slashed incident triage time by 45 %; planners receive natural-language summaries instead of raw sensor codes.
User satisfaction ↑ 30 pp: revamped Angular dashboard and chat-style assistant replaced legacy SOAP portal.
Release cadence doubled — CI/CD shrank average time-to-feature from monthly drops to bi-weekly sprints.
Cloud-agnostic & future-proof: micro-services run identically on Azure AKS or on-prem Rancher, meeting Swiss data-residency rules while keeping a path open to full public-cloud scale. Together these gains position the client as an innovation leader in European smart-freight logistics.