AI for DevOps
Senior DevOps Engineer with a strong background in CICD and Observability and Monitoring and skilled in tools like Elasticsearch, Docker, Kubernetes,Terraform, and Ansible. I focus on automating using DevOps tools or scripting using shell and python.
Traditional AI:
AI has been around since the 1900s.
Primary use case of traditional AI: predictions
Example: climate prediction
Models are fed with historical data from 1970, 1971, and so on.
DevOps usecases:
Predict events,patterns related to incident management.
loganalysis
Aws autoscaling
Generative AI:
Primary use case of traditional AI: Generate content
Can be text, images, and videos
Input → GenAI → generates a new image without copying from the internet
Devops Usecases:
ask new kubernetes manifest —→ LLM —> generates manifest according to requirement
LLM(Large Language Model):Focused on text
AIOps = Traditionals AI + Generative AI
How LLMs are powerfull?
llama:407B(trained with 407 billion parameters) :supercomputers with huge Gpus and Tpus of memory they are capable of parallel computing and super fast
Prompt Engineering:
the more relevant information provided as input to LLM gives better output
input:prompt
DevOps Landscape:
Git
CI/CD
Infrastructure as Code
Configuration management etc
AI Landscape for DevOps:
AI chatbots: chatgpt, deepseek, clod, llama
AI agents: github copilot (free now), bolt.new (not free)
AI assistants: github copilot, pieces for developers
Programming language: FastAPI, flast
(python interacts with models locally making api calls)