Senior Machine Learning Engineer · Berlin · 2017–Present

Mira Kovač

Senior Machine Learning Engineer · Berlin, Germany

[ CV · 2026 · 6 YEARS IN MACHINE LEARNING ]

Experience

3 positions · 2017 – 2026 · Berlin
2024 — Present

Senior Machine Learning Engineer

Nimbus Systems · Berlin, Germany

  • Built a two-stage re-ranking pipeline for product recommendations, serving 2M+ monthly active users while reducing session churn by 18%.
  • Led the model-serving migration to GPU-backed inference, increasing throughput 3x with accuracy within 0.5% of the legacy baseline.
  • Owned the end-to-end MLOps stack from training orchestration to canary deployment, supporting 120+ model variants with 99.9% training uptime.
2020 — 2024

Machine Learning Engineer

Helix AI · Berlin, Germany

  • Designed time-series forecasting models for supply-chain logistics, cutting predicted-demand errors by 22% compared to legacy models across 18 warehouses.
  • Built a feature-store service powering 15+ internal data pipelines, cutting feature engineering time by 40% for ML teams across the company.
  • Co-led the migration of the legacy model-serving platform to a GPU inference stack, reducing cold-start latency by 60%.
2017 — 2020

Data Scientist / Machine Learning Engineer

Brightwave Analytics · Berlin, Germany

  • Developed click-through prediction models that improved campaign ROI by 25% across three ad platforms at scale.
  • Built and maintained ETL pipelines for a 200GB daily telemetry dataset, plus evaluation suites for 8+ model families.
  • Collaborated with product teams to ship the first production ML feature in the user onboarding flow.

Skills

7 skills · machine learning focus
Machine Learning MLOps PyTorch Computer Vision NLP & Foundation Models Time-Series Forecasting Experiment Tracking

Side Projects

3 projects · open-source & internal tools
01

Edge LLM Inference Runner

Lightweight GPU inference wrapper for local model serving, cutting cold-start latency by 60% in edge deployments.

02

Tabular Predict Dashboard

A self-serve feature-preview and model-comparison dashboard for internal ML teams, adopted by 60+ engineers across 4 teams.

03

Time-Series Simulator

A synthetic data generator for training forecasting models without access to proprietary historical datasets.

Publication

1 publication · 2023
2023

Efficient Serving for Multi-Model ML Pipelines

ML Systems Workshop · Berlin — co-authored on scalability practices for multi-model serving at scale.

Contact

Email mira.kovač@nimbus.io
Location Berlin, Germany
Available For new roles — senior ML engineering & MLOps positions