Nenad Nikolic

Development of reproducible data workflows used to retrieve and transform data used for training and fine-tuning LLMs, effectively turning DevOps approach into MLOps. Evaluation of Kubeflow, Airflow and Flyte. Used Flyte for implementing data workflows with Python. Integration with and evaluation of several LLMs. Custom training data processing logic on the GCP (Google Cloud Platform; Google Kubernetes Engine and Cloud Functions).

Backend software development with Java, Kotlin, Spring Boot, Kafka, MongoDB with DevOps/DevSecOps activities on the Kubernetes platform running on top of Azure cloud (incl. Azure IoT), GitHub Actions for CI/CD pipelines, Flux for GitOps. API design and implementation. Security scanning (dependencies & code): Trivy, Dependabot, CoPilot, SonarQube. Implementation of electricity usage prediction models (based on ONNX models and Tablesaw library) for optimizing energy consumption and thermal comfort. Frontend development with TypeScript, React and Vite TypeScript. Prediction PoC and data engineering using Python and Pandas. Scripting using Bash and Python. Monitoring with Prometheus, Grafana, ELK. Confluence and Jira.

Development and maintenance of backend and frontend components for the otto.de online shop with robust and comprehensive testing ensuring maximum uptime. Designed and developed backend applications using Scala and Spring Boot, integrating with MongoDB for data management and Kafka for product updates. Created frontend components with AngularJS, TypeScript and Freemarker templates, ensuring seamless interaction with backend systems via RESTful APIs and HAL responses. Thorough OpenAPI-based API documentation in the central API portal. Conducted detailed analysis of product requirements, specifying and implementing new features based on them. Interactive API testing with Insomnia. Implemented extensive testing including unit (with ScalaTest) and integration tests (with WireMock and TestNG), smoke tests, layout tests (using Playwright and Backstop), and consumer-driven contract testing. Infrastructure managed via generated Jenkins pipelines in the Amazon/AWS cloud, deploying workloads as scalable EC2 instances, also using SQS, S3 buckets, Secrets Manager, and Lambda functions, monitored/observed via Micrometer library, CloudWatch and Grafana. MongoDB Atlas for dev and live databases. Gradle was used as the build tool. Used libraries like Dropwizard, Cats, Cats Effect, Quicklens, and many more. Trunk-based development with feature flags. Used Backstage internal developer portal, incl. for API documentation.