01 / CONTEXT
Practical AI needs reproducible evaluation.
An AI competition doesn't just check a final answer. Participants work with different datasets, notebooks, models and metrics, and the results must be evaluated under comparable and safe conditions.
The official platform supports the preparation and stages of the competition through problems, submissions, datasets, rankings and automated evaluators adapted to each test.
02 / INT80 ROLE
Technical Committee and Software Engineering.
Liviu Popa contributes to the technical committee, evaluates and mentors participants, and works on the software component of the competition, including the automated judging infrastructure.
- 01
Platform development
Contribute to functionality used for competitions, resources and evaluation.
- 02
Evaluation infrastructure
Flows for submission, controlled run, metrics and reproducible results.
- 03
Technical commission
Validation of the technical format and support for the conduct of tests.
- 04
Mentoring and assessment
Feedback for participants and link between educational objective and implementation.
03 / PLATFORM MAP
From problem to verifiable score.
Problems
Statement, data, rules and metric defined for each sample.
Artifacts
Predictions, code or notebooks received in a controlled format.
Judge
Automatic assessment adapted to the type of task and constraints.
Result
Consistently published scores and rankings for the stage.
04 / RESULT
Infrastructure that lets the competition be about AI.
The platform reduces manual work around assessment and provides participants with a consistent environment for preparation and competition. The technical component remains flexible enough for tasks with different formats and metrics.