Calibrating retrieval confidence for low-resource catalog search
CURRICULUM VITAE / 2026
Mira Kovač
Senior Machine Learning Engineer · Berlin
I build reliable retrieval and ranking systems for products used at scale — with a focus on measurable quality, fresh signals, and calm operations.
Get in touch
01
Experience
-
2023 — NOWFULL-TIME
Senior Machine Learning Engineer
SoundCloud · Berlin
- Led a two-tower retrieval refresh for track discovery; improved qualified plays per session by 5.4% in a controlled experiment.
- Moved freshness features to an hourly pipeline and added segment-level drift alerts, cutting stale recommendations for new releases.
-
2020 — 2023FULL-TIME
Senior Machine Learning Engineer
Delivery Hero · Berlin
- Built restaurant and dish retrieval models using query, menu, and order signals; raised NDCG@10 by 8% across German-language searches.
- Introduced time-based offline replay and city-level holdouts, catching seasonal ranking regressions before rollout.
-
2017 — 2020FULL-TIME
Machine Learning Engineer
Zalando · Berlin
- Trained learning-to-rank models for fashion search from click and purchase events, with guardrails for position bias and sparse queries.
- Added image embeddings to candidate retrieval, improving cold-start coverage for new catalog items without increasing p95 latency.
02
Skills
- Python
- PyTorch
- Ranking & retrieval
- Recommender systems
- Feature engineering
- Offline evaluation
- SQL
- ML observability
03
Selected projects
04
Publication
05
Contact
For thoughtful problems in search, ranking, and applied ML.