# MLOps objeves

Objectives focused on machine learning operations.
***Team focused on MLOps.***
***Team focused on MLOps Workflow/pipeline challenges.***

## ML Workflow synnary

### ML Workflow



### ML Workflow devied by phases

1. Problem definition
- KPIs definitions
3. Data phase
- Data sourcing/Data collection
- Data exploration
- Data preparation
- Data segregation
4. Learning phase
- Model Training
- Model evaluation
5. Operationalization (o16n) phase: ML model available in production env
- Model deployment
- Model monitoring

## ML pipeline

ML pipeline = Continuous and iterable ML Workflow improvement

## Challenges

1. Problem definition (data analyst challenges)
- Well-defined glosary
3. Data (data scientist challenges + data data engineer challenges)
- Docs
- Well defined tech-stack
- Secutiry staregy
- Data governance
- Standard sources
- Data versioning
- Data audit
4. Learning phase (data data engineer challenges = machine learning engineer challenges)
- Docs
- Well defined tech-stack
- Testing staregy
6. Operationalization (o16n) phase (Developer + DevOps challenges)
- apps and Infra Documentation
- Testing staregy
- Well defined tech-stack
- Secutiry staregy
- Security audits Implementation
- Semantic Versioning approach
- Continuous methodologies
- CI
- CD/CD
- DevOps adobtion
- IaC adoption
- GitOps implementation
- Infra/apps Observability strategy (tool)
- Logs Strategy (standar)
- Metrics startegy (standar and Per line)
- Actions startegy (standar and Per line)
- Alert startegy (standar)

## Contution

MLOps = Culture to automate, manage, and audit ML workflow (DevOps + ML workflow)

Team focused on MLOps Workflow/pipeline challenges