CS-01BANKING
Kafka & OpenSearch as-a-service on an internal marketplace
At a major European bank, internal teams were installing their streaming and search stacks by hand. We designed and shipped Kafka and OpenSearch/Elasticsearch as-a-service products on the group's internal marketplace — a full self-service lifecycle for internal client teams.
What we did- +Productized the full self-service lifecycle: cluster creation, service accounts, scaling, extra storage, upgrades/migrations, S3 backups
- +Chose OpenSearch to keep Elastic Platinum licensing costs under control
- +Automated monitoring and alerting with Prometheus exporters and Dynatrace on Kubernetes
- +Administered clusters to audit and security standards: ILM, snapshots/restore, index templates, encryption, ACLs
- +Automated operations with Ansible and Python; resolved complex production incidents
StackKafkaOpenSearchElasticsearchKubernetesAnsiblePythonPrometheusDynatraceLogstash
CS-02LUXURY / RETAIL
Streaming on AWS & managed search on Elastic Cloud
At a global luxury group, we ran the Kafka platform on AWS and the Elasticsearch estate on Elastic Cloud (managed) behind public-facing and internal data flows.
What we did- +Administered public and internal Kafka clusters on AWS: creation, scaling, encryption, ACLs, performance and HA
- +Administered Elasticsearch deployments on Elastic Cloud (managed): Logstash, ingest pipelines, ILM, snapshot/restore, migrations
- +Built real-time pipelines with Kafka Streams and Kafka Connect, integrating Oracle and GCP
- +Elasticsearch transforms (pivots) feeding analytics
- +Automated the infrastructure with Terraform on AWS (S3, ECS, EKS) and Python
StackKafkaKafka StreamsKafka ConnectElasticsearchElastic CloudAWSTerraformKubernetesPythonPrometheus
CS-03MEDIA INTELLIGENCE
Semantic search & real-time alerting for media monitoring
For a media-monitoring provider, search is the product: users save queries and expect matching content the moment it is ingested.
What we did- +Deployed Elasticsearch clusters on Kubernetes (EKS) and optimized query performance
- +Built embedding-based semantic search on Elastic, alongside lexical search
- +Added an AI layer on top of retrieval
- +Used percolation to match incoming content against saved searches and feed results to users right after ingestion — real-time alerting
StackElasticsearchKubernetes (EKS)Vector & semantic searchEmbeddingsPercolatorPythonAI/LLM
CS-04SUPPORT AUTOMATION
AI agents automating day-to-day support operations
Recurring support tasks were eating the team's days. We deployed AI agents to take them over end to end.
What we did- +Deployed task-automation agents on daily support workflows
- +Close to 80% of tasks now handled automatically, end to end
- +Remaining and edge cases routed to humans
StackAI agentsLLMAutomation
CS-05TELECOM
5G database observability & ML anomaly detection
At a global telecom equipment manufacturer, we built the supervision platform for 5G database equipment, used by engineers across France, India and Finland.
What we did- +Deployed the platform with Ansible: 5 Elasticsearch clusters, a Hadoop cluster (12 DataNodes / 3 NameNodes), Kafka, plus collectors
- +Built a Lambda architecture over logs and KPIs with Python and Logstash
- +Industrialized ML anomaly detection on logs — log-template mining (Drain/SLCT), event matrix, PCA, density clustering (DBSCAN/HDBSCAN/OPTICS) — with PySpark and Spark Streaming reading from Kafka
- +Jenkins jobs to start and stop equipment monitoring; ILM policies on the clusters
- +Earlier phase: a KPI-monitoring PoC (Logstash/Elasticsearch/Kibana) and a chaos-testing tool whose failure scenarios joined the product's CI regression campaign
StackElasticsearchKafkaHadoopPySparkAnsibleJenkinsPythonMachine Learning