Founding ML Engineer

## About LIR Labs We are building the autopilot for industrial biology, using proprietary hardware and deep learning to optimise the >2% of global electricity consumed by water infrastructure. By replacing lagging chemical sensors with continuous real-time imaging, we prevent costly process upsets and eliminate energy waste at scale. ## About The Role We are looking for a founding engineer for our office in Hannover to own our ML stack from first principles. You will inherit a working computer vision baseline and decide what to keep, what to replace, and what to build next. As your scope matures, you will extend into the modelling layer that connects vision-derived features to downstream process variables, laying the foundation. You will work directly with the CTO and have full visibility into technical and commercial decisions from day one. We prioritize willingness and ability to upskill quickly over existing skills and motivation to make a difference and deploy fast are key. ## Responsibilities - Own the computer vision architecture: evaluate the existing baseline, make architectural decisions, and take full accountability for model performance in production - Build and maintain cloud inference serving live testing sites across Europe - Design and maintain pipelines that ingest and process live sensor data from industrial installations - Build the correlation layer between vision derived features and downstream process variables - Structure model outputs as clean, documented inputs for the operator-facing interface - Collaborate with the founding team to integrate operational observations with model behaviour ## Qualifications - 3+ years in applied ML (not necessarily professional), with at least one deployment of a computer vision system - Fluent in German and English - A clear opinion about segmentation architectures and the ability to defend it - Experience extracting performance from small datasets: augmentation, transfer learning, semi-supervised approaches - Solid time series modelling skills: feature engineering from sensor data, handling irregular sampling, autocorrelation and non-stationarity - Suf