ANIMAS 6615

Benefits of ANIMAS OnPrem AI

• The customer has total control over the hardware infrastructure, data content and localization, configuration, and security, with no required access to third-party resources (e.g., Large Language Model [LLM] providers or third party services requiring an API key).

OnPrem AI Planning & Analytics

• Integrates with diverse data sources and destinations
• Supports fine-grained access control and data encryption
• Real-time data ingestion, transformation, and processing
• Scales horizontally to handle large data volumes
• Language translation (i.e., Korean and Japanese)

ZIngest Data Management

ETL

Harvest

Conditioning

Unstructured Data Sources

• Multimedia (Audio/Video)
• Sensor Data
• Communications
• Word / PowerPoint

Semi-structured Data Sources

• XLM / JSON
• NoSQL
• Email
• Reports

Structured Data Sources

• Relational Databases
• Proprietary Databases
• Spreadsheets, emails, documents, reports, presentations

Form Factor Highlights

Ground-breaking compute capability (64 cores and 2TB of RAM) in a highly compact platform footprint, making it the ideal platform for high-performance computing at the edge. Available in an Air Transportation Association (ATA) carry-on compliant case.
Can be scaled to accommodate additional components and technology.

Fits in an ATA compliant transit case engineered by Tracewell.

Maximum Configuration

• Processors: 64 cores
• Memory: 2TB
• Storage: 10, 2.5” drives
  (8 SATA/SAS or 10 NVMe)
• PCIe Slots: 2 (capable of holding Nvidia L4 GPU or equivalent form factor GPUs)
• SYSTEM: 18”D, 13” W, 3.47”H

Note: AMD EPYCTM processor

Performance Highlights

• Support for 20 concurrent users with latencies under 10 seconds for large context queries (average of 20 tokens)
• Throughput of 300 queries per minute (QPMs) for 10 concurrent users
• Average ingest time for PDF documents: 170 pages/minute

Solution Hardware

• We have Subject Matter Experts that can assist with full spectrum hardware & software solutions; for tactical & strategic solutions.
• Benchmarking and validation conducted on industrial-grade architectures with Nvidia GPU accelerators (H100 and L40S)
• Capable of supporting customer requirements from small Tactical-Edge deployments to large-scale enterprises 

Zapata Technology’s ZIngest Features

• ZERO Trust Data Encryption at Rest and in Transit – MPO-approved process ensures the security of data both at rest and during transit, enabling both encryption and decryption
• All Purpose Data Processing Management – All types of data, structured, semi-structured, and unstructured, are processed and moved among diverse data sources and systems
• Visualized Data Processing – The complete cycle of data processing pipelines including design, maintenance and monitoring is managed in a web-based GUI
• High Throughput, Data Provenance, Reliability, & Security –
  Schedule-based data stream processing pipelines are distributed, prioritized, and tracked which assures delivery
• High level of modulization supported by a large library of readily used components, expression language, and scripting extension increases pipeline development productivity and adaptability to change

FEDDATA’s OnPrem-GPT AI Features

• Enterprise-level RAG AI for multi-user environments, on-prem solution with no external resources needed, supporting thousands of concurrent users on multiple nodes
• Proven and customizable LLMs (e.g., Mistral, Mixtral), with a bring-your-own LLM option available
• Data retrieval augmented with reference sources and document images, with AI responses filtered using relevance scores from semantic search
• Advanced text embedding models and vector store databases with fine-tuning capabilities for specific use cases, and AI data management for large volumes of data
• Separate data collections with ad-hoc user access policies, supporting various data formats (PDF, Microsoft Suite, raw text, web scraping)
• Integration capabilities with other data sources and prompt engineering and guard-railing for accurate context-specific answers