From real-time Kafka ingestion through ElasticSearch and PostgreSQL storage, Spark and Hadoop processing, Grafana and Kibana dashboards, to HPC and GPU-driven ML workloads. Built on serious deployments — SK broadband CDN monitoring, KISTI's HPC platform, the Statistics Training Center's big-data learning environment.
A data platform earns its value only when ingestion, storage, processing, visualization, and ML interlock cleanly. Build them separately and operations get heavy — change one and the others wobble.
We've built every layer directly — SK broadband's CDN monitoring (Kafka + ELK), KISTI's HPC and GPU-driven ML platform (EDISON), the Statistics Training Center's Virtual Learning Environment.
Combined with our own products (ElasticUpload for ingestion, JMeter-based load verification), we deliver a solution owned end to end, from initial build through to live operations.
Apache Kafka-based real-time log and event ingestion. Hundreds of servers stream data through losslessly to downstream systems.
ElasticSearch, PostgreSQL, InfluxDB — the right store for the workload. Search, aggregation, and time-series all live on the same platform.
Hadoop, Spark, and Hive for big-data processing through to ML training on HPC and GPU clusters. Kubernetes and Slurm automate resource provisioning.
Grafana and Kibana bring time-series, search, and resource usage onto one screen. OpenTelemetry standards link in any observability tooling.
JMeter-based load scenarios verify the lossless aggregation and accuracy of ingestion pipelines before they go live.
On-premise, AWS, Naver Cloud, Kakao Cloud — environments don't constrain the architecture. We tune to each.
Service
Real-time ingestion, storage, and visualization pipelines.
→
Service
HPC, GPU, and Kubernetes-based ML workload operations.
→
Service
Distributed load verification for lossless and accurate pipelines.
→
Product
100GB+ ingestion channel for large data uploads.
→
We built a platform that aggregates logs from hundreds of servers and monitors system resource and network usage. A Kafka + ELK design pushed resource utilization to the limit, with JMeter-based load scenarios verifying lossless aggregation.
KISTI's HPC-based ML learning platform automates high-end infrastructure such as GPU-driven ML model execution.
A VM-based big-data learning environment that provisions per-learner setups in an automated way for tailored learning.
SK Inc. AX