- Vishakha Sadhwani
- Posts
- DevOps to Forward Deployed Engineer: A Free 7-Role Learning Ladder
DevOps to Forward Deployed Engineer: A Free 7-Role Learning Ladder
A step-by-step path from your first Linux command to deploying AI agents for customers, using only free resources.
Most roadmaps treat tech roles as separate tracks. In practice they stack: each role is the one before it plus one or two new skills.
1. DevOps Engineer = Linux + Scripting + CI/CD + Containers
2. Cloud Engineer = DevOps + Cloud
3. SRE = Cloud Engineer + Observability + Incident Response
4. Platform Engineer = SRE + Kubernetes + IaC
5. AI Infra Engineer = Platform Engineer + GPUs + Model Serving
6. AI Solutions Architect = AI Infra + LLMs/RAG + Customer Architecture
7. Forward Deployed Engineer = Agentic Harness + Production Code + On-site Delivery
Nothing you learn on the way up is wasted. Below are free resources for every new skill, including crash courses, plus one project per role. Each project builds on the last.
1. DevOps Engineer
Linux + Scripting + CI/CD + Containers
You build the path code takes from a developer's commit to a running service, and you make that path fast, safe and predictable.
Linux
Scripting
CI/CD
Containers
Project: Containerise a small app and ship it with a CI/CD pipeline. Follow the Docker CI/CD Project. For an overview first, watch All DevOps Concepts in 25 Minutes. For daily practice, use #90DaysOfDevOps.
2. Cloud Engineer
DevOps + Cloud
You run everything from role 1 on cloud infrastructure: compute, networking, identity and storage. You also keep an eye on cost.
Project: Deploy your role 1 app to a cloud account inside a VPC you built yourself. Put it behind a load balancer and use least-privilege IAM. Set a budget alert before you start.
3. Site Reliability Engineer
Cloud Engineer + Observability + Incident Response
You keep production running, define what "reliable enough" means, and turn every outage into a lasting fix.
Observability
SLOs
Incident Response
Project: Add a Prometheus and Grafana dashboard to your role 2 deployment and define one SLO. Then break the deployment on purpose and run the incident as the incident commander. Finish with a blameless postmortem.
4. Platform Engineer
SRE + Kubernetes + IaC
You build the internal platform that other teams deploy onto. Your users are your own developers.
Kubernetes
IaC
Platform Design
Project: Rebuild your stack from roles 1–3 in Terraform: the network, the cluster and the monitoring. Deploy the app to Kubernetes through GitOps. Then destroy everything and recreate it with one command.
5. AI Infrastructure Engineer
Platform Engineer + GPUs + Model Serving
The platform stays the same, but the workload changes. You keep GPU clusters well utilised and turn model checkpoints into fast, scalable endpoints.
Project: Work through the End-to-End DevOps + AIOps Project. Then serve a small open model with vLLM on your role 4 cluster. Track tokens per second and time to first token in Grafana, and load-test until it fails.
No GPU? Free Google Colab and Kaggle notebooks are enough for CUDA basics.
6. AI Solutions Architect
AI Infra + LLMs/RAG + Customer Architecture
You turn a customer's "we want AI" into a design that is accurate, secure, affordable and fits their existing stack.
LLMs
RAG
Customer Architecture
Project: Watch Full-Stack GenAI. Then write a one-page architecture for a fictional customer, for example: "RAG over 2 million internal PDFs, must stay in their VPC, $5k/month budget." Include a diagram, the trade-offs, a cost estimate and the first three questions you would ask the customer.
7. Forward Deployed Engineer
Agentic Harness + Production Code + On-site Delivery
You don't stop at the design. You build the system inside the customer's environment, often working alongside their team, and make sure it works from day one.
Agentic Harness
Production Code
On-site Delivery
Project: Build your role 6 design. Use the Agent SDK for the agent and connect it to one real tool over MCP. Wrap it in a FastAPI service and deploy it on your role 4 platform. Add evals and a dashboard, then write a handoff doc the customer's team could run with on their own.
Crash Course Quick List
Topic | Link |
|---|---|
DevOps Concepts | |
Linux | |
Git | |
CI/CD | |
GitOps | |
Docker | |
Docker CI/CD Project | |
Networking | |
VPC Hands-On | |
VPC Flow Logs: Part 1 | |
VPC Flow Logs: Part 2 | |
AWS Roadmap | |
GCP Roadmap | |
Azure Roadmap | |
JWT vs OAuth | |
Prometheus (playlist) | https://www.youtube.com/playlist?list=PLy7NrYWoggjxCF3av5JKwyG7FFF9eLeL4 |
Grafana (playlist) | https://www.youtube.com/playlist?list=PLyJqGMYm0vnO9osZ-EBV6iu2l10muE2A- |
Kubernetes: Part 1 | |
Kubernetes: Part 2 | |
Kubernetes: Security | |
Terraform | |
Terraform Project | |
Terraform Project Repo | |
vLLM on Kubernetes | |
DevOps + AIOps Project (playlist) | https://www.youtube.com/playlist?list=PLXkUFcIv0_b7rzZe0o_2-GOS2qn5-0OQy |
System Design | |
APIs | |
Caching and Redis | |
Vector Databases | |
RAG | |
Full-Stack GenAI | |
Tool Calling | |
MCP and AI Agents | |
KodeKloud: Linux | https://kodekloud.com/courses/crash-course-linux-for-absolute-beginners |
KodeKloud: Docker | https://kodekloud.com/courses/crash-course-docker-for-absolute-beginner |
KodeKloud: Kubernetes | https://kodekloud.com/courses/crash-course-kubernetes-for-absolute-beginners |
KodeKloud: AWS Basics | |
KodeKloud: AI-Powered DevOps | https://kodekloud.com/courses/crash-course-ai-powered-devops |
KodeKloud: MCP | https://kodekloud.com/courses/crash-course-mcp-for-beginners |
How to Use This
Treat this as a ladder, not a menu. You don't need to reach role 7 to get hired: roles 1 and 2 alone qualify you for real jobs.
Before moving up, build the project for the role you're on. Each project extends the previous one, so by the end you'll have a single system rather than seven toy projects. It starts as a Dockerfile and ends as a cloud-hosted, monitored, Terraform-managed AI agent running on GPUs.
Every resource listed is free. Course availability and URLs change, so if a link has moved, search for the title.