DR. ATABAK KH
Cloud Platform Modernization Architect specializing in transforming legacy systems into reliable, observable, and cost-efficient Cloud platforms.
Certified: Google Professional Cloud Architect, AWS Solutions Architect, MapR Cluster Administrator
Welcome to my collection of research, thoughts, and experiences on leadership, cloud computing, DevOps, and software engineering. Personal site — not affiliated with any employer. Disclaimer
A static profile says what a user has liked, searched, viewed or bought. A sequence says how that interest evolved, and how recently. After a LinkedIn feed-ranking conversation, I ran...
Pipelines still ran. Dashboards still loaded. Nobody deleted anything, fixed definitions, or admitted the mart nobody used cost thousands a month. That was a graveyard - not a platform.
The demo worked. The pilot dashboard looked green. Then production asked for SLAs, cost predictability, audit trails, and the same answer on Tuesday that you got on Monday.
Everyone is talking about AI. The real question is whether your team builds systems that last - or relies on a few heroes, knowledge nobody wrote down, and demos that...
The promotion did not come with a new toolchain. It came with ambiguity, stakeholders who disagree, and problems that do not fit in an IDE.
Governance decks showed up after the first incident. By then the architecture had already decided what agents could read, what was logged, and what could not be undone.
When I was doing my PhD, I spent years building computational models for cancer detection. It sounded glamorous from the outside - algorithms, prediction, science. In reality it was a...
Right-Time AI: Small Copilots, Big Proof - An End-to-End PoC
Most AI projects fail before the model is the problem. They fail because the platform underneath cannot answer freshness, ownership, lineage, or trust. This is the scorecard I keep coming...
Problem: Either you pay too much (min too high) or you get tail latency (min too low). Balance it.
Many teams scale only on average CPU. That is easy, but it may miss queues and tail latency. Use demand signals such as queue depth or request rate to scale,...
Takeaway: For robust GO prediction, start with homology + PLM baselines, add label smoothing on PPI, and only then graduate to GNNs/multimodal fusion.
By mid-2025 “agents” were on every vendor slide. What separated a demo from something operators could live with was rarely a better system prompt. It was permissions, logging, approval flows,...
Problem. Independent multilabel classifiers for Gene Ontology (GO) often violate the ancestor rule: if a child term is predicted “on,” all its ancestors must also be on. Curators then fix...
Product APIs have SLOs. Data pipelines often only have “it ran green”. That gap is why dashboards lied, models trained on stale data, and on-call found out from Slack.
Goal: page for user pain, not random metric spikes. Burn-rate alerts do that by measuring how fast you’re spending the error budget for your SLO.
How to design a multi-service sync pipeline from cloud to on-premises (or between systems) using hexagonal architecture, Cloud Workflows, and event-driven orchestration.
Goal: Make results re-runnable and comparable (CAFA-style).
Takeaway: You don’t need perfection-just a directionally correct cost per request/job.
Context: BigQuery is fast to adopt-and easy to overspend on. Here’s a comprehensive checklist I use in migrations to avoid common cost pitfalls.
Takeaway: a handful of Terraform patterns prevent surprise spend, reduce pager incidents, and make audits easy.
Engineers need stable schemas. Business needs trustworthy definitions. Without something written down, both sides guess - and every dashboard, migration, and AI pilot pays for it.
Purpose: A short checklist to avoid inflated or unstable GO results.
Idea: One smoothing step over a normalized PPI graph can yield consistent gains before you build a full GNN.
By mid-2024 almost every team had a GenAI demo. The ones that stalled when they touched real data did not fail on prompt engineering. They failed on stale tables, fuzzy...
Takeaway: Frozen PLM embeddings + linear classifier = strong, fast baseline for GO prediction.
Takeaway: You can cut cost and improve reliability/observability without any access to PII or raw logs. Here’s the artifact-only method I use.
Deadlines do not remove trade-offs. They hide them until production. Under pressure I learned to make trade-offs visible, written down, and reversible when I could.
How to design Spark pipelines for real-time derived data and ML scoring; with the right shuffle, skew, and storage choices. This article use real-time campaign segments as the example throughout....
Migration plans talk about tools and timelines. What broke after go-live, in my experience, was usually schema drift, partitioning assumptions, batch dependencies, cost models, and expectations nobody wrote down.
I’ve seen migrations fail in planning docs and committee meetings, long before anyone started writing the actual code. The first problem is almost never the technology.
I keep seeing programs celebrate “we are on AWS / Azure / Alibaba” while the way people work with data barely changed. Landing a VM or a dump into the...
how to design and tune Spark pipelines for high-volume lead, view, and sales event data; covering shuffle behavior, skew handling, and storage layout so pipelines stay correct and scalable.
Basic ideal solutions
Cloud Migration in a simple way
Digital Transformation with Agile + DevOps Culture
Finance sees the invoice. Architecture creates it. The durable cost improvements I saw did not come from better spreadsheets - they came from changing how data was partitioned, scheduled, and...
I became a lead because I could unblock anything under my responsibility. Then I became the bottleneck for everything while the team waited.
Accountability on Resilience Engineering
Most of the time, dotnet developers setup and configure a windows server and then occupy the entire server for a single project or a micro service project using web services....
Apache Spark is an engine for Big Data processing. One can run Spark on distributed mode on the cluster. In the cluster, there is master and n number of workers....
Before going thorugh configuration and build of kubernetes packages, we should install minikube or Kubernetes. My primary goal was to try Kubernetes on VMs the simplest way. I have chosen...
Docker containers and services do not even need to be aware that they are deployed on Docker, or whether their peers are also Docker workloads or not. Whether your Docker...
What is Docker Compose
This document shares an experience on setting up Kubernetes and then configuring your .NET Core application into your cluster. It will also help automate your deployment using a Kubernetes cluster...
This document would share an experience on setting up dockerized master-slave hadoop and spark on top of them. Then config your environment and listen to steaming data. It will also...
WHAT IS PROTEIN FUNCTION? The meaning of organic capacity is vague, and the correct significance of the term fluctuates in view of the setting in which it is utilized. Clearly...
Showing 7 of 49 papers
Transforming legacy systems to cloud-native architectures
Building reliable systems with proper monitoring and alerting
Data-driven approaches to reduce cloud spending
Dr. Atabak Kh is a cloud platform engineer specializing in modernization, observability, and cost optimization. His research focuses on practical approaches to improving system reliability and reducing cloud costs through data-driven methods.
Current areas of interest include p95-driven autoscaling, privacy-first cloud audits, and SLO-based alerting systems.
This is a personal blog. The views, thoughts, and opinions expressed here are my own and do not represent, reflect, or constitute the views, policies, or positions of any employer, university, client, or organization I am associated with or have been associated with.