Senior ML Engineer (Routing) - Team *instinctools

*instinctools is a software development company that provides custom software solutions for businesses of all sizes. Our team works closely with clients to understand their specific needs and provide personalized solutions that meet their business requirements.

*instinctools is looking for a Senior ML Engineer (Routing) for one of our clients.

Location: Europe, Poland, Kazakhstan

The platform allows users to book rides and other services, including ride-hailing, intercity travel, delivery, and task assistance. It operates in more than 700 cities across 47 countries and is the second most downloaded mobility app in the world.

You will work on the machine learning systems behind ETA accuracy and routing quality at scale. The project combines deep learning, large-scale geospatial data processing, traffic estimation, and low-latency production model serving.

Your work will directly influence how accurately the platform predicts arrival times, adapts to changing traffic conditions, and supports routing decisions. Model improvements are evaluated not only through offline metrics but also through online experiments, with measurable effects on ETA accuracy, cancellations, and business outcomes.

What you will get hands-on with

• ETA prediction at scale: Design and improve machine learning models that refine routing engine estimates, from gradient-boosted trees to neural networks and Transformer-based architectures.

• Geospatial data engineering: Process large-scale GPS traces, perform map-matching, and build spatial aggregations that turn raw movement data into reliable modeling inputs.

• Traffic estimation: Develop road-level speed estimates and historical traffic profiles to support time-of-day-aware ETA predictions.

• Routing quality optimization: Improve ETA calculation, smoothing, and rerouting logic across changing traffic conditions.

• Experiment-driven ML: Evaluate model performance using offline metrics and online experiments, including shadow tests and switchback designs.

• Production ML systems: Take models from development to low-latency serving, with attention to latency, throughput, monitoring, and model lifecycle management.

• End-to-end ML ownership: Work across data analysis, feature engineering, model development, deployment, monitoring, and retraining.

This role combines applied machine learning research with production engineering. You will need to understand not only whether a model performs well offline, but also whether it improves real-world routing outcomes when deployed.

What you will do

• Design, build, and optimize ML models that correct and refine ETA estimates generated by the routing engine.

• Develop traffic-estimation models using large-scale GPS data to generate road-level speeds and historical traffic profiles.

• Implement map-matching approaches that associate noisy GPS traces with the road graph and develop spatial aggregations for movement and speed data.

• Improve ETA calculation, smoothing, and rerouting logic to reduce the gap between predicted and actual arrival times.

• Translate business and routing objectives into ML problems, selecting appropriate proxy metrics, loss functions, and non-functional requirements.

• Account for asymmetric prediction errors when underestimating and overestimating arrival times carry different costs.

• Lead model evaluation from offline accuracy and routing-quality metrics through online experiments, including shadow tests and interference-aware designs such as switchbacks.

• Partner with backend engineers to deploy models into production serving infrastructure and meet latency and throughput requirements.

• Collaborate with product and operations teams to translate routing and traffic analysis into concrete features and requirements.

• Help extend ETA capabilities across different service verticals.

• Own the production ML lifecycle, including model serving, monitoring for data and concept drift, and developing retraining pipelines as traffic patterns, cities, and map data evolve.

What we expect from you

• 5+ years of experience in machine learning engineering, including at least 3 years building and deploying deep learning models in production.

• Direct experience building regression, forecasting, or other supervised ML systems for real-world prediction problems.

• Expert-level proficiency in Python and its core data science libraries, including Pandas, NumPy, and Scikit-learn.

• Strong SQL skills and experience working with distributed data processing technologies such as PySpark.

• Practical experience with PyTorch and gradient-boosting libraries such as CatBoost, XGBoost, or LightGBM.

• Ability to design an ML system from scratch in at least one domain, covering data analysis, processing, feature engineering, model development, and production serving.

• Experience translating business goals into ML objectives, selecting appropriate proxy metrics and non-functional requirements, and designing or substantially contributing to experiments that demonstrate real-world impact.

• Experience using MLOps tools and practices to manage the machine learning lifecycle.

• Experience deploying models to production ML serving infrastructure and optimizing inference latency.

• Understanding of concept drift and practical approaches to detecting and managing it.

• Experience working with large-scale geospatial and behavioral data, such as GPS traces, and spatial indexing techniques such as H3.

• Experience with distributed data processing and the ability to build reliable ML pipelines for production workloads.

Nice to have

• Academic background in Computer Science, Mathematics, or another quantitative discipline.

• Domain expertise in ETA or travel-time prediction, traffic estimation, or routing engine quality.

• Hands-on experience with open-source routing engines such as Valhalla, OSRM, or GraphHopper.

• Understanding of map-matching, speed profiles, road-graph tiles, and historical traffic data.

• Experience building mapping, location-based, or geospatial products.

• Experience developing solutions for markets where map coverage and address data are limited or unreliable.

• Experience with cloud data and ML platforms such as BigQuery or Databricks.

• Relevant cloud platform certifications.

• Experience with distributed deep learning training.

What we offer

• A technically challenging ML project. Work on production machine learning systems where model quality directly affects routing accuracy and real-world business outcomes.

• End-to-end engineering ownership. Contribute across the full ML lifecycle, from geospatial data processing and model development to production deployment, evaluation, and retraining.

• Large-scale geospatial challenges. Work with GPS traces, traffic estimation, road networks, and location-based data in an international product environment.

• Experiment-driven development. Validate improvements through offline evaluation and online experiments rather than relying on model metrics alone.

• A professional and ambitious team. Collaborate with engineers and specialists working on complex machine learning and routing problems.

• Professional development. Access learning opportunities, seminars, conferences, and time to explore new technologies.

• Language learning support. Co-funding for English language courses.

Join *instinctools to build production-grade machine learning systems that improve ETA accuracy, routing quality, and mobility experiences at scale.

Our expectations of the ideal candidate:

• 5 or more years of machine learning engineering experience, with at least three of them building and deploying deep learning models in production
• Direct experience building regression, forecasting, or other supervised ML systems for real-world prediction problems.
• Expert-level proficiency in Python and its core data science libraries and SQL (e.g., PySpark, Pandas, NumPy, Scikit-learn, PyTorch; gradient-boosting libraries such as CatBoost/XGBoost/LightGBM)
• The ability to design an ML system from scratch in at least one area, including data analysis, processing, and feature engineering through to a model serving in production
• Experience turning a business goal into an ML problem with the right proxy metrics and non-functional requirements, and designing or substantially contributing to the experiments and statistical evaluation that prove impact on real-world behavior
• Experience using MLOps tools and practices to manage the ML model lifecycle
• Experience deploying models to production on ML serving infrastructure and optimizing for latency, and awareness of concept drift and how to detect and manage it
• Comfort working with large-scale geospatial and behavioral data (e.g., GPS traces, H3 spatial indexing) and distributed processing.

Nice to have:

• An academic background in a quantitative field such as Computer Science, Mathematics, or a related discipline will be a plus
• Subject matter depth in ETA / travel-time prediction, traffic estimation, or routing-engine quality
• Hands-on experience with open-source routing engines (e.g., Valhalla, OSRM, GraphHopper) and concepts such as map-matching, speed profiles, road-graph tiles, and historical traffic
• Experience in mapping, location, or geospatial products
• Experience building for developing markets, where the underlying map and address data is weak
• Experience with a cloud data and ML platform such as BigQuery or Databricks (certifications a plus), and with distributed deep-learning training

English: Upper-Intermediate 

Soft skills:

• The ability to influence teammates and partner teams, and to communicate complex results clearly

Important Notice Regarding Your Personal Data Security

Thank you for your attention and cooperation.

share Margarita Lukyanenka Recruiter We will be glad to see you in our team!