MLOps Engineer

Companion.energy; Oost-Vlaanderen; Hybrid; EUR 60,000 - 90,000; Full time

Job summary

Companion.energy is seeking an experienced MLOps Engineer to build the production-ready infrastructure for forecasting and control models across hundreds of assets. You’ll own deployment, backtesting at scale, and real-time monitoring of algorithms that drive millisecond decisions in energy markets.

The role emphasizes end-to-end MLOps—from packaging and versioning to monitoring and alerting—within a fast-growing startup focused on energy optimization and risk management.

Qualifications

• Strong Python software engineering and maintainable code practices.
• Experience taking ML models to production with deployment, monitoring, and rollback.
• Fluent in MLOps toolchain: experiment tracking and model registry (MLflow or equivalent), orchestration (Dagster, Airflow, Prefect).
• Familiar with containers, Kubernetes, CI/CD, and infrastructure as code.
• Experience with drift detection, data quality, and model-performance monitoring (Evidently, Great Expectations, Soda).
• Time series as a core problem: late/out-of-order data, backfills, and careful train/test separation.
• SQL skills and robust data pipelines; streaming or near-real-time systems (Kafka).
• Knowledge of energy systems: electricity markets, load forecasting, asset dispatch, balancing mechanisms.
• Previous startup/scale-up experience.

Responsibilities

• Package, deploy, version, and rollback forecasting and control models; ensure reproducible training and a clear promotion path to production.
• Monitor models in real time feeding market decisions with stringent latency requirements.
• Operate cross-asset deployment across hundreds of assets with data quality considerations.
• Ensure time-series correctness, handling irregular intervals, gaps, and backfills for backtests.
• Build scalable backtesting infrastructure replaying historical data across portfolios.
• Optimize real-time performance as intraday trading expands.

Skills

Python

MLOps

MLflow

Dagster

Airflow

Kubernetes

CI/CD

Infra as Code

Prometheus

Grafana

Evidently

Great Expectations

Soda

Time Series

SQL

Kafka

Feast

TimescaleDB

PostgreSQL

Energy markets

Startup experience

Tools

MLflow

Dagster

Airflow

Kubernetes

Docker

Prometheus

Grafana

Feast

TimescaleDB

PostgreSQL

Kafka

Job description

We help large enterprises cut energy costs and valorize flexibility by steering their energy use in sync with market prices and renewable generation.

We're hiring an MLOps Engineer to build the infrastructure that gets our forecasting and control models into production and keeps them working there: across hundreds of assets, on time series data, against live markets.

You’ll own model deployment, backtesting at scale, and the real-time monitoring of algorithms that drive second‑by‑second decisions worth millions.

About Companion.energy

Companion.energy connects the financial side of energy management (contracts, markets, risk) with the operational side (assets, processes, sites). We model complex energy contracts and flexible assets, forecast demand and production, and translate predictions into automated, second‑by‑second control decisions that move megawatts and money. Our software is used by large B2B enterprises and energy players to lower OPEX, maximize revenue, increase renewable usage, and manage risk.

Why Companion.energy?

• Real stakes: our models steer real assets against live market prices. When a model degrades, it costs money the same day.
• Foundations: MLflow, orchestration, and models already earning money in production. What’s missing is the layer above it: the one that makes shipping a model routine and keeps hundreds of them honest in real time. That layer is yours to design.
• Monitoring that is actually hard, and interesting because of it: real‑time constraints because markets don’t wait, and hundreds of assets each with their own data quirks.
• A mission that matters: better forecasting and smarter control accelerate the transition to a flexible, renewable grid.

What you’ll work on

• The path from model to production: packaging, deployment, versioning, and rollback of forecasting and control models, so that shipping a model change is routine rather than an event. Model registry, reproducible training runs, and a clear promotion path from experiment to production.
• Monitoring in real time: our models feed market decisions that cannot wait for tomorrow’s report.
• Monitoring at scale: we run models across potentially hundreds of assets, each with its own behaviour and its own data quality problems.
• Time series correctness: everything we do is time series. Late and out‑of‑order data, irregular intervals, gaps and backfills, and point‑in‑time correctness so that a backtest cannot accidentally see the future.
• Backtesting infrastructure that runs at scale: replaying years of market and asset data across portfolios of assets and markets.
• Real‑time performance: as we move into intraday trading and add markets and customers, latency and throughput become product features rather than implementation details.

We’re looking for someone who builds and operates production systems, and who takes satisfaction in making other people’s models fast, reliable, and observable.

What we are looking for

• Strong software engineering in Python: you write code other people maintain, and you have opinions about tests, CI, and rollback.
• You have taken machine learning models to production and operated them there: deployment, versioning, monitoring, and being on the hook when they break.
• Fluent in the standard MLOps toolchain, and clear about when not to reach for it: experiment tracking and model registry (MLflow or equivalent), orchestration (Dagster, Airflow, Prefect), containers and Kubernetes, CI/CD, infrastructure as code, and metrics and alerting (Prometheus, Grafana).
• You have built drift, data quality, and model performance monitoring (Evidently, Great Expectations, Soda or similar) and tuned the alerting so that it stays useful at scale.
• Time series as a first‑class problem: late and out‑of‑order data, irregular intervals, resampling, backfills, and why a naive train/test split leaks the future.
• SQL, and the discipline to keep scheduled pipelines healthy rather than merely running.
• Streaming or near‑real‑time systems (Kafka or equivalent).
• Feature stores (Feast or equivalent) and serving models under latency constraints.
• Time series stores at volume (TimescaleDB, PostgreSQL, or similar).
• Knowledge of energy systems: electricity markets, load forecasting, asset dispatch, balancing mechanisms, or related domains.
• Previous working experience in a fast growing start up or scale up.

Location

Ghent, Belgium (hybrid)

How we hire

Intro call - technical case - follow-up conversations with the team and founders.

We keep it simple and can move fast.