Machine Learning Engineer

About Realytics:

We’re building the first foundation model that predicts brick-and-mortar performance from brand, location, service quality, and assortment. Our model helps retailers improve existing stores and pick the next best sites with calibrated, explainable predictions.

We train on 200M POIs worldwide, 6B customer opinions and selective finance data.

We work with customers from North America, EU and APAC.

What you’ll do:

• Frame success metrics (traffic/revenue proxy, location ranking, calibration) and own E2E modeling.
• Design and ship a multimodal model across geo, tabular, multilingual text and images in future.
• Reconcile micro-level store predictions with grid-level spend and aggregates.
• With help of our engineers build the data pipeline: feature store, training snapshots, backtests, batch/online inference.
• Set up robust validation, uncertainty & reporting.

Data you’ll work with:

• POI facts and history.
• User feedback: ratings + multilingual reviews.
• Financials data & industry ratings.
• Geo layers: demographics, transport/isochrones
• Satellite images

What we’re looking for (must-have):

• 5+ years in ML/DS shipping production models; strong stats & modeling fundamentals.
• Experience with Tabular ML and NLP. Forecasting, ranking and spation-temporal as an advantage.
• Practical rigor with noisy proxies, leakage/causal hygiene, calibration & interpretability.
• Practical understanding of optimisation of data processing and model fitting.
• Comfortable with Python, Catboost/XGBoost, PyTorch/TF, SQL, Spark/Dask; building real pipelines and orchestration.

Nice to have:

• Geo engineering (H3/quadkey, OSM, isochrones)
• Payments data experience.

Tech stack:

Python, PyTorch, CatBoost/XGBoost, Spark/Dask, SQL/Postgres(+PostGIS), H3, Airflow, Docker/K8s, MLflow/W&B, cloud (GCP/AWS/Azure).

Location and compensation

We are remote first company (mostly UTC - UTC+4).

We could help with relocation to Cyprus

We suggest competitive salary + equity.