Skip to content
About

About Rokibul Hasan

I'm a senior software engineer in Dhaka, Bangladesh, working across two areas that keep meeting each other: AI systems and the infrastructure that keeps them running. Day to day that means Go and Python services, Kubernetes controllers, model serving, and the observability needed to know whether any of it is healthy.

The short version

I joined AppsCode Inc. as a software engineer in January 2023 and was promoted to senior software engineer in January 2025. Throughout, the work has been ACE, AppsCode's container engine. Most of my work there sits in the multi-cluster layer: a dashboard that manages public and private Kubernetes clusters from one place, Kubernetes operators and APIs that automate authentication and authorization with secure kubeconfig generation, and a multi-tenant monitoring system built on OpenTelemetry, Thanos, ClickHouse and Perses. I also packaged ACE as a self-hosted product so it can be installed on both vendor-managed and self-managed clusters.

In parallel I've worked with clients through Upwork since 2021, where I hold "Top Rated Plus" status and a 100% Job Success Score. That work has ranged from Go and Python backends to GitOps automation on enterprise OpenShift, and, most recently, to real-time AI systems: a production voice agent with retrieval, voice cloning and a lip-synced video avatar, and before it the backend for a game where players hold a spoken conversation with an LLM-powered character.

Before that I studied Computer Science & Engineering at East West University, where my thesis on cloud data security was published in IEEE Xplore.

How I work

A pattern shows up in most of what I've built: the interesting problem is rarely the feature itself, it's what happens when there are many of something. Many clusters, many tenants, many concurrent players hitting one model server. That's usually where the design decisions live.

  • Read the upstream code. When I depend on a project, I end up in its source, which is how most of my open-source contributions started.
  • Make state explicit. Kubernetes controllers and asynchronous job pipelines both reward the same habit: write the desired state down, then reconcile toward it.
  • Instrument before optimizing. Metrics, logs and traces first; guesses about the slow part are usually wrong.
  • Automate the second time. Ansible, ArgoCD, Helm and GitHub Actions exist so that the third deployment looks exactly like the first.

Education & research

B.Sc. in Computer Science & Engineering, East West University, 2019 – 2023 (CGPA 3.49/4.0). My coursework leaned toward distributed systems, operating systems, databases and networks, which turned out to be the right preparation for cluster work.

My thesis, "Data Security in Cloud Computing using a Hybrid Algorithm Approach", proposed a hybrid approach to strengthening data security in cloud environments. It was published in IEEE Xplore: read the paper on IEEE Xplore.

Competitive programming

I solved over 1,000 problems across online judges and took part in more than 200 contests: Codeforces (max rating 1695), CodeChef (max 1848), LeetCode (max 1609) and a 6-star gold badge on HackerRank. It's where I learned to reason about complexity before writing code, a habit that still pays off when a query or a reconcile loop is doing far more work than it needs to.

See the full work history

Where to go next