Gopala Krishna Akkineni

Home About

From patching servers to running ML platforms.

I’ve spent over four years making infrastructure boring in the best way: repeatable, watched, and easy to undo. Here’s how I got here and what I care about.

It started with shell scripts

My first job, in July 2020, was as a Linux administrator at Ignitarium in Kochi. The work was hands-on and unglamorous: patching kernels, hardening servers, setting up cron jobs and log rotation, and helping developers figure out why their app was slow on a box I was responsible for.

The part I liked most was writing shell scripts. User provisioning, file cleanup, process monitoring, backups. Anything I did twice by hand, I tried to script the third time. That habit is still the core of how I work.

Moving into the cloud

In 2021 I joined Taal Tech in Bengaluru as a Cloud/DevOps engineer. For the first time I was working across AWS, Azure and GCP, provisioning instances, virtual machines and storage, managing IAM roles and security groups, and rotating keys. I contributed Terraform and CloudFormation templates so environments were consistent and version-controlled. I also helped set up Jenkins and GitHub Actions pipelines and learned to troubleshoot Kubernetes when pods kept restarting.

Studying while working

I finished my diploma in Electronics and Communication Engineering in April 2020 and went straight to work. At the same time I started a B.Tech in the same field and did it alongside my jobs at Ignitarium, Taal Tech and Amistad. I finished it in April 2023. That is why the dates on my résumé overlap: the degree and the jobs ran side by side.

There is also a gap on the résumé, between Taal Tech, which ended in February 2022, and Amistad, which started in May 2022. In that stretch my semester exams and my project submission landed at the same time, so I took a couple of months off to finish them. Then I went back to work.

Learning what production really means

At Amistad Technologies in Hyderabad, from 2022 to 2023, I was closer to the fire. I handled L1/L2 support for AWS services and networking incidents and kept SLA adherence at 95%. That job taught me what matters when something breaks: a clear alert, a tested backup, a rollback you trust. I helped run disaster recovery tests on backups and DNS failover, and wrote reusable Terraform modules that cut manual provisioning in half.

Going deeper

Next I did my Master of Science in Management Information Systems at the University of Illinois Springfield, finishing in December 2024. It gave me the business side of technology: why a platform exists, who pays for it, and how to explain a technical tradeoff to people who don’t live in a terminal.

Where I am now

Since March 2025 I’ve been a DevOps engineer working with State Farm in Bloomington. I lead the design of multi-cloud infrastructure for AI and ML across AWS and GCP: CI/CD pipelines, EKS and GKE clusters, GPU and TPU capacity for training, secrets, monitoring and incident response. I work closely with data scientists and ML engineers, and some of what we’ve shipped together are LLM-based apps running in production.

It’s the most interesting mix I’ve worked on. A model is only useful once it’s reliably in front of real users, and that last mile is exactly where DevOps lives.

What I care about

The ideas I keep coming back to, whatever the stack.

Do it by hand once, then automate it

The first time, I want to understand the steps. The second time, it becomes a script, a Terraform module or a pipeline job. Manual work is where drift and 2 a.m. surprises come from.

Rollback should be cheaper than debugging

Small changes, canary releases and GitOps mean a bad deploy is a quick revert, not an all-hands investigation. Fix forward when you understand the problem, roll back when you don’t.

If it isn’t monitored, it isn’t done

Dashboards and alerts ship with the service, not in a follow-up ticket. And an alert should mean someone needs to act. If it doesn’t, it’s noise and it trains people to ignore the real ones.

Security belongs in the pipeline

A scan that fails the build is worth more than a report nobody reads. Secrets live in Vault or a secret manager, access follows least privilege, and credentials rotate on a schedule.

Cost is a reliability problem too

Idle GPUs and forgotten instances are money that could fund real work. Autoscaling and scheduled shutdowns are part of running a healthy platform.

Write it down

Runbooks, onboarding notes and honest root cause reviews. The next person on call, sometimes me, should not have to rediscover what we already learned.

Tools

Tap one to see where I’ve used it.

Cloud

Infrastructure as code

CI/CD and GitOps

Containers

Observability

Security

ML platforms

Scripting and collaboration

Pick a tool above.

Education

Electronics first, then the business side of technology. I did the B.Tech alongside my jobs, so the dates overlap. "Studying while working" in my story above has the details.

  1. Master of Science, Management Information Systems

    University of Illinois Springfield, Springfield, IL · Dec 2024

  2. Bachelor of Technology, Electronics and Communication Engineering

    Aditya University, Surampalem, Andhra Pradesh, India · Apr 2023

  3. Diploma, Electronics and Communication Engineering

    AANM & VVRSR Polytechnic, Gudlavalleru, Andhra Pradesh, India · Apr 2020

Hiring for DevOps or platform work?

I’m open to DevOps, SRE and platform engineering roles. Email is the fastest way to reach me.