HRIA: Romanian Artificial Intelligence Hub
AI Research Engineer, National University of Science and Technology POLITEHNICA Bucharest
From research prototype to production: training and evaluating ML/DL models, then building the CI/CD, infrastructure and MLOps workflows that run them reliably within Romania's national AI hub.
Problem
Research teams produce promising models, but turning experiments into reproducible, deployable, monitored systems needs engineering discipline that research pipelines usually lack.
Solution
A combined research + engineering role covering the complete AI lifecycle: model development and optimisation on one side, CI/CD, compute provisioning, containerisation and MLOps standards on the other.
My contribution
- Design, train and evaluate machine-learning and deep-learning models; run experiments and iterate on architectures against research objectives.
- Profile and optimise algorithms for accuracy, speed and resource efficiency, turning prototypes into clean, production-quality code.
- Build CI/CD pipelines for ML workloads: automated tests, model validation, versioning and release workflows.
- Provision and manage training and inference environments: containerisation, orchestration, monitoring and reproducibility across environments.
- Standardise data handling, experiment tracking and model deployment (MLOps), working alongside researchers and engineers.
Stack
Python
Machine learning
TensorFlow
PyTorch
MLOps
Docker
Kubernetes
GitLab CI/CD
Prometheus
GrafanaLinks
Context
HRIA (Hubul Român de Inteligență Artificială) is a national AI research and development initiative hosted at POLITEHNICA Bucharest. The role spans both sides of the research-to-production gap.
Research
- Model design, training and evaluation; experiment analysis and architecture iteration.
- Algorithm optimisation: profiling for accuracy, latency and resource use.
Engineering
- CI/CD for ML: automated testing, model validation, versioning and release workflows.
- Compute environments for training and inference, containerised and orchestrated, with monitoring and reproducibility guarantees.
- MLOps standards for data handling, experiment tracking and deployment.
