Building production-grade, zero-trust cloud infrastructure on GCP.

Google Certified Cloud Engineer Google Certified Cloud Engineer Badge Click to verify on Credly specializing in Infrastructure-as-Code, CI/CD, and serverless data pipelines. I design and build secure, scalable, and cost-efficient systems that solve real-world problems.

Status: Actively seeking full-time Cloud/Data Engineering roles.
Get in Touch

Projects

A selection of technical case studies demonstrating enterprise-grade practices.

End-to-End DevSecOps Pipeline

An automated pipeline that enforces supply chain security by building, scanning, and deploying only cryptographically signed and vulnerability-free container images.

Cloud BuildBinary AuthorizationArtifact RegistryDevSecOps
View Case Study →

Advanced GKE Cost Optimization

A deep-dive into combining HPA, VPA, and Cluster Autoscaler to build a cost-effective, resilient GKE cluster that dynamically adapts to workload demands.

GKEHPAVPACluster AutoscalerCost Optimization
View Case Study →

Intelligent Cloud Operations Agent

An AI agent using LangChain and a self-hosted LLM to provide natural language control over GCP operations and infrastructure.

LangChainTerraformFastAPIGCP
View Case Study →

Real-Time Serverless Data Pipeline

A high-throughput, serverless data pipeline on GCP for ingesting, validating, and analyzing IoT sensor data in real-time.

Pub/SubCloud FunctionsBigQuery
View Case Study →

Zero-Touch CI/CD Pipeline

A fully automated CI/CD pipeline using GitHub Actions to build, test, and deploy a containerized application to Google Cloud Run.

GitHub ActionsTerraformCloud Run
View Case Study →

Production-Grade Service Mesh

An implementation of a service mesh on GKE using Istio to enforce zero-trust security, canary deployments, and rich observability.

GKEIstioTerraformObservability
View Case Study →

Experience

Data Engineer

Vodafone (Remote)

  • Engineered and optimized event-driven, serverless data pipelines using Cloud Functions and Pub/Sub, processing over 1,000 events per second with sub-2-second latency.
  • Automated the provisioning of all GCP infrastructure using Terraform, enforcing strict module versioning and achieving 100% IaC coverage for the production environment.
  • Re-architected a multi-terabyte BigQuery data warehouse by implementing table partitioning and clustering, reducing query costs by over 60% and improving dashboard load times by 200%.

AI & ML Researcher

Association for the Advancement of Science and Technology in BiH

  • Researched and implemented computer vision models for object detection tasks.
  • Published findings in a peer-reviewed academic journal.

GCP Mentor

Google Developers Group Sarajevo

  • Presented a technical deep-dive on "Automated Testing in DevOps Pipelines" to 50+ attendees.
  • Mentored junior developers on GCP architecture, Terraform best practices, and cloud-native application design.

Before you go...

Let's connect. I'm always open to discussing new projects and opportunities.

Email LinkedIn

Intelligent Cloud Operations Agent

LangChainPythonTerraformFastAPI

Problem Statement

Cloud operations teams rely on manual queries for routine diagnostics. This project aimed to create an AI agent to automate these tasks using natural language, providing instant Infrastructure-as-Code (IaC) awareness.

Architecture

User Interface(Gradio)FastAPI Backend(on GCE VM)LangChainAgentSelf-Hosted LLMCloud StorageRead StateFirestoreCRUDCloud LoggingQuery

Technical Decisions & Outcomes

  • Architectural Choice: A self-hosted Hugging Face model on a GCE VM was chosen over proprietary APIs to ensure data privacy and eliminate API costs, accepting the trade-off of increased management overhead.
  • Quantifiable Outcome: Automated 5 core operational checks, reducing the time for routine diagnostics from 5-10 minutes manually to under 15 seconds via natural language.
View on GitHub

Real-Time Serverless Data Pipeline

Pub/SubCloud FunctionsBigQueryPython

Problem Statement

IoT devices generate high-volume data streams that require immediate ingestion, validation, and analysis. This project's goal was to build a highly scalable, low-latency, and cost-effective serverless pipeline to handle this workflow.

Architecture

IoT DevicesData SourcePub/SubIngestionCloud FunctionValidationBigQueryAnalysisPub/SubAlerts Topic

Technical Decisions & Outcomes

  • Architectural Choice: Cloud Functions Gen 2 was selected for its superior event-driven capabilities and longer processing timeouts, making it ideal for complex validation logic without needing a containerized environment like Cloud Run.
  • Quantifiable Outcome: The pipeline was architected to handle 1,000+ messages per second with an average end-to-end latency of under 2 seconds.
View on GitHub

Zero-Touch CI/CD Pipeline

GitHub ActionsTerraformDockerCloud Run

Problem Statement

Manual deployments are slow, error-prone, and unscalable. This project aimed to create a fully automated CI/CD pipeline to achieve zero-touch build, test, and deployment of a containerized application to a serverless platform.

Architecture

Developergit pushGitHub Actions CIBuild, Test, PushArtifact RegistryGitHub Actions CDTerraform Plan & ApplyCloud RunDeployed Service

Technical Decisions & Outcomes

  • Architectural Choice: Google Cloud Run was the ideal deployment target over GKE due to its fully managed, serverless nature, which eliminated Kubernetes management overhead.
  • Quantifiable Outcome: The pipeline fully automated the deployment lifecycle, reducing the process from 30+ minutes to an average execution time of 4 minutes and 15 seconds.
-85% Reduction in Deployment Time
< 5 min From Push to Live
View on GitHub

End-to-End DevSecOps Pipeline

Cloud BuildBinary AuthorizationArtifact RegistryDevSecOpsSecurity Scanning

Problem Statement

In modern software delivery, the risk of deploying compromised or vulnerable container images is significant. The objective was to build a zero-trust CI/CD pipeline that cryptographically verifies the integrity and security of every image before it can be deployed to production.

Architecture

Developergit pushCloud BuildBuild & ScanScan Fails: BlockCloud BuildSign Image w/KMS AttestorArtifact RegistryProduction RepoCloud Run

Technical Decisions & Outcomes

  • Architectural Choice: Using Binary Authorization with a Cloud KMS-based attestor was the critical security decision. This provides a non-repudiable, auditable cryptographic signature for each image, which is a much stronger guarantee than simply using image tags.
  • Quantifiable Outcome: The pipeline automatically blocked a test deployment containing 3 critical CVEs, proving its effectiveness as a security gate. The secure, end-to-end process from `git push` to a signed, production-ready artifact was fully automated and took an average of 6 minutes.
View on GitHub

Advanced GKE Cost Optimization

GKEHPAVPACluster AutoscalerCost Optimization

Problem Statement

Production Kubernetes clusters often suffer from either wasteful over-provisioning or performance-impacting under-provisioning. The challenge was to create a multi-layered autoscaling strategy that automatically right-sizes both pods and infrastructure, minimizing cost without sacrificing availability.

Architecture

GKE ClusterNode Pool(e.g., e2-standard-4)PodVPA (Vertical)HPA (Horizontal)Cluster AutoscalerPod-level scaling (HPA/VPA) drives Node-level scaling (CA)

Technical Decisions & Outcomes

  • Architectural Choice: Implementing "Pause Pods" for overprovisioning was a key strategy. These low-priority pods reserve buffer capacity, allowing high-priority workloads to be scheduled instantly during a scale-up event, rather than waiting for new nodes to be provisioned.
  • Quantifiable Outcome: By using Pause Pods, the cluster's response time to a sudden demand spike was reduced from 3-5 minutes (waiting for new nodes) to less than 1 minute, dramatically improving application availability while still benefiting from scale-to-zero cost savings during idle periods.
View on GitHub