Software

Building Industrial AI Pipelines For Reliability, Scalability And Cost Efficiency Advanced AI enables deep insights for battery manufacturing, by discarding the paradigm of single modal, single task, and training on limited data to a pattern of multimodal, multitask, and pre-training on massive data 1. But high cost of hardware and staff training hinders adoption 2.

GOAL FACTORY FOUNDATION MODEL

WHAT DO WE NEED FOR FACTORY DEPLOYMENT?

Real time micro-defect image analysis

Reliability: Ensures AI systems consistently perform as expected in factory operations, even under varying conditions or workloads.

Real time macro-defect image analysis

Scalability/Availability: Enables AI systems handle increasing data volumes or compute demands and guarantees that AI systems are operational and accessible whenever required for critical factory processes. 3

LLM enabled Out of Control Action Plan OCAP

LLM enhanced PM Preventative Maintenance)

Root cause analysis

HOW IS IT DONE?- Example Method

Heterogeneous Computing: Leverages diverse hardware (e.g., CPUs, GPUs, FPGAs) to optimize the performance of different types and sizes of models. Model Lifecycle Management: Forces model and data traceability to ensure continued relevance in the face of rapid product and process innovation.

Sources: 1 Zhang, Large Scale Foundation Models for Intelligent Manufacturing Applications: A Survey, Zhang et al.; 2 “High Cost Hinders AI Adoption Among IT Clientsˮ, Consultia; 3 Edge AI A Taxonomy, Systematic Review and Future Directions, Jill et al.

AI SUITE

EDGE INFERENCE AI Training

ModelOps

Model Pruning Model Quantization Edgebox Configuration

Extracted from the PDF; charts and images may read imperfectly.