Improving LLM Accuracy & Performance

Gen AI is driving transformative changes in the ways we do business, with a wide range of use cases being tested and applied in all departments and functions across the enterprise. To keep costs low and efficiency high, many organizations are now fine-tuning existing LLMs, to customize them for a specific organization and use case.

Thus, much attention is being given to:

• Improving the accuracy of LLMs
• Increasing LLM performance
• Keeping costs low
• De-risking the model by adding guardrails / training on proprietary data alone

This MLOps Live session shares best practices and pragmatic advice on successfully improving the accuracy and performance of LLMs while mitigating challenges like risks and escalating costs. Practical examples were shared including techniques to overcome common challenges using tools such as Databricks Mosaic AI and their new open LLM, DBRX.