What problems should enterprises actually trust an LLM to solve?
summary
Flip AI builds a Devops LLM for root-cause analysis that interrogates observability stacks (CloudWatch, Datadog, Splunk) and returns hypotheses with corroborating metrics surfaced and data-source provenance. Models start from open-source LLaMA-class foundations, are instruction-tuned on chaos-engineered fault datasets of logs/traces/metrics, and target tens-of-billions-parameter scale so they fit on two GPUs for single-tenant on-prem or VPC deployment. Claims: zero-shot useful RCAs on unseen customer data, avoids multi-tenant training, and sidesteps $2–3M custom-training costs.
AI-written summary · about
chapters (7)
- 0:00Introduction and speaker background
- 3:03Why most AI POCs fail in enterprises
- 8:22Key requirements for enterprise LLM adoption
- 13:18Trust and transparency in LLM outputs
- 18:56Training custom models vs. open source
- 24:11Enterprise deployment challenges
- 28:42Choosing the right LLM use cases
related talks
42mOpen Source LLMs — Powering AI Accessibility and Innovation
47mA Step-by-Step Guide to Securing Large Language Models (LLMs)
30mSecuring AI Workloads: Building Zero-Trust Architecture for LLM Applications
21mAI-Powered O11y with LLMs: Resilience in Cloud PaaS Archs
01h 49mAI engineering with open access LLMs that lie, curse, and steal
28mAnyscale's Unified Platform for LLM Development and Deployment
12mKeynote: How LLMs Have Changed the Way Organizations Do AI
20mSecure, private LLMs in your cloud with Anyscale Endpoints