// mira kovač

Mira Kovač

Senior Machine Learning Engineer — Berlin

ML engineer focused on retrieval, ranking, and evaluation systems for large models and high-stakes predictions. Berlin-based, EU time zones.

Email hello@mirakovac.dev GitHub github.com/mirakovac LinkedIn linkedin.com/in/mirakovac Location Berlin, DE

// experience

2017

DeepMind

London / Remote

Machine Learning Engineer — 2017 → 2019

  • Built neural policy-search pipeline for partially observable control tasks, reducing sample complexity versus prior baselines.
  • Designed offline reinforcement-learning curation workflow with reward-noise filters, stabilizing long-horizon training.

2019

Spotify

Berlin

Machine Learning Engineer — 2019 → 2022

  • Led two-stage podcast retrieval and re-ranking model, improving candidate diversity and click-through.
  • Shipped offline-online evaluation split for ranking candidates, preventing rollout of metric regressions.

2022

Bloomberg

Berlin

Senior Machine Learning Engineer — 2022 → Present

  • Architected tabular foundation-model fine-tuning pipeline for entity risk signals, improving forecast calibration.
  • Deployed fraud-signal classifier with review queue, cutting false positives while preserving auditability.
  • Mentored four engineers and ran ML lifecycle reproducibility guild.

// skills

// side projects

// publication

Kovač, M., Ruiz, A., & Okafor, T. — “Sparse Reward Shaping via Curriculum Generation”

NeurIPS 2023 Workshop on Reinforcement Learning. Spotlight.