Data Exists But AI Can't Use It Directly.
With scattered data locations and formats, RAG, model training, and analytics reporting each require separate work. Ultimately, AX projects fail to transition from PoC to operations.
Data Lakehouse for AI Utilization
AkashiQ
AX's Bottleneck is
Not Model,
But Data.
When adopting AI, actual enterprise data is scattered across DB, Hadoop, file servers, log repositories, quality documents, and business systems. Security prevents exporting to external clouds. AkashiQ organizes data within internal networks and transforms it into structures usable by analytics and AI.
When Organized in Internal Networks, AX Operates.
AkashiQ connects storage, catalog, format conversion, unified queries, and AI utilization as one flow without external data export.
Within Secure Network,
Build Data Lakehouse.
Even where external data export is restricted, AkashiQ connects internal S3 storage, catalog, analytics engine, and AI workloads.
Designed assuming data, queries, and analysis results don't go out through external cloud APIs.
Enterprise Internal Data
Collect and organize manufacturing equipment, quality documents, batch records, logs, DB, and Hadoop data by internal network standards.
MES · EDMS · LIMS · QMSHadoop/ HDFS · DB · File ServerLogs · Documents · Sensor Data
→
AkashiQ Internal Data Lakehouse
Deploy S3 storage, Parquet/Iceberg, Catalog, permissions, and analytics engine in internal networks to create data utilization flow.
MinIO S3/ Ceph S3Parquet · Iceberg · Metadata CatalogTrino · SparkAccess Policy · Audit Log · Lifecycle
✕
Block External Cloud Dependency
Configure data difficult to transmit externally per security policy so internal AI and analytics services can use it.
Minimize / Block External LLM APIsInternal Registry · Repository OperationsSAMANDA · IKIN · BI · Model Training Integration
Beyond Storage,
To AX Data Hub
AkashiQ is not simple object storage. It's an enterprise data lakehouse stably storing large-scale data, managed with metadata and permissions, connected to analytics engines and AI workloads.
Air-Gapped S3 Data Storage Structure
Stably store large-scale data within internal networks and provide S3 storage structure suitable for AI utilization.
MinIO and Ceph-based object storageBlock/File/Object integrated considerationInternal network installation and operations standard design
Catalog, Permission, and Lifecycle Management
Create data hub structure spanning collection, refinement, transformation, and sharing, managing data based on metadata and catalogs.
Parquet and Iceberg table structureCatalog-based data asset managementDomain-specific data separation and access policies
Analytics and AI Utilization Optimization
Integrate with Trino, Spark, Presto for SQL-based analysis of multiple data sources, providing data foundation for RAG and model training.
Distributed SQL analytics engine integrationRAG and model training data foundationKosmosAI product suite integration
Large-Scale Data Operations
in Practical Structure
Built.
Design data lakehouse as operationally feasible infrastructure. Consider storage, network, catalog, analytics engine, migration, DR, permissions, and audit together.
Data Lakehouse Unified Management
Connect S3 storage, data catalog, analytics engine, and data format in single flow, managing from storage to utilization.
S3 BucketCatalogParquetIceberg
Hadoop to S3 Migration Architecture
Support data replication, consistency validation, phased cut-over, and rollback strategies for migrating existing Hadoop/HDFS to S3-based lakehouse.
HDFSDistCpDual-writeRollback
Secure Operations and Governance
Include authentication, permissions, audit logs, deployment, rollback, and monitoring in operational processes.
Access PolicyAudit LogLifecycleAir-Gapped
AI Workload Utilization Foundation
Connect lakehouse data to RAG, model training, reporting, and data services to enhance enterprise AI utilization.
RAGModel TrainingSQL Analytics
In Manufacturing and Pharma AX,
Explained by Data Foundation Layer.
MANUFACTURING AX
AAS and Equipment Data Utilization Foundation
Collect equipment, sensor, process, and status data in standardized structure, creating data foundation for AI analysis and operational decisions.
PHARMA AX
GMP Environment Data Operations Support
Manage batch records, quality documents, change history, and log data in internal network, providing structure linkable to permission, audit, and history policies.
KOSMOSAI LINK
Expansion with SAMANDA and IKIN
Data organized in AkashiQ can flow to SAMANDA's operational decisions, IKIN's RAG search, BI/reporting, and model training data.
Data Operations Method
Changes with AkashiQ.
Before
After with AkashiQ
BeforeData ScatteredHadoop, DB, file servers, and log repositories separated, making data utilization paths complex.
AfterInternal Network S3 HubConsolidate large-scale data into internal network S3-based lakehouse, organizing storage and utilization flow.
BeforeExternal Export RestrictionsSecurity policy makes it difficult to send data to external cloud or external AI APIs.
AfterAir-Gapped AnalyticsConnect catalog, analytics engine, and AI workloads in internal environment to reduce external dependencies.
BeforeAnalytics DisconnectionStorage, catalog, and analytics engine operated separately, making connection to AI utilization difficult.
AfterUnified QueryUnified SQL-based analysis of multiple data sources with Trino, Spark foundation.
BeforePost-PoC StagnationEven when the PoC succeeds, data preparation is manual every time, so it never transitions into operations.
AfterAI Ready DataProvide AX data foundation flowing to RAG, model training, reporting, and data services.
AkashiQ FOR DATA FOUNDATION
Without External Export,
Build AX Data Foundation.
AkashiQ transforms large-scale data into S3-centric data lakehouse,
designing AX foundation flowing to analytics and AI utilization.