Location: Poland - Warsaw
DT (NASDAQ: APPS) is unifying the full mobile experience. Our growth platform powers the mobile ecosystem by bringing together advertisers, app developers, and device partners, to create a unique and premium value exchange, that results in a personalized and seamless mobile device experience.
At DT, we move fast, think big, and build collaboratively. We believe great ideas can come from anywhere, and we trust our teams to take ownership, challenge convention, and make an impact from day one.
Here, you’ll have the opportunity to grow your career while contributing to products and experiences used by millions around the world.
With us, you will enjoy competitive compensation, equity awards, and a comprehensive benefits package designed to invest in your total well-being, all within a hybrid and flexible work environment.
The Role
Digital Turbine is expanding its machine-learning capabilities across real-time advertising systems. As the number and complexity of our models grow, we need an architecture that enables Data Scientists to experiment productively while delivering models into production with predictable performance, scalability, and cost.
As a Lead ML Systems Engineer, you will define the technical direction for our ML productionization framework. Your focus will span the full path from feature extraction and curation through model productionization and inference, shaping how data, features, and models move through the system and how each stage can be optimized for performance, scalability, and total cost of ownership.
You will guide the transition from Databricks-based experimentation to production-ready artifacts using the most appropriate technologies for each workload, including SQL based processing, native Python, Polars/Ray, distributed Spark processing, and optimized CPU/GPU execution approaches, including ONNX-based models and technologies such as nvForest.
You will act as the critical bridge between our ML Platform and the engineers behind core runtime - rAVE, our in-house real-time inference runtime. A key part of this role is developing architectural approaches that reduce duplicated computation on the ML side, package models efficiently, and establish strict contracts so models run predictably when handed off to the inference environment.
About the Lead ML Systems Engineer Role
Own the ML productionization architecture: Define the technical direction for reusable frameworks that move models from experimentation into production, covering model dependencies, feature processing, execution environments, validation, and deployment.
Bridge experimentation and production: Design efficient and reproducible workflows connecting Databricks-based model development with production execution, including training artifacts, featurization pipelines, feature consistency, and dependency management.
Optimize feature extraction and curation: Define approaches for efficiently extracting, transforming, joining, and curating features across analytical and production data systems. Evaluate data access patterns, processing strategies, and dependencies to reduce redundant computation, unnecessary data movement, and infrastructure costs.
Optimize model execution: Determine how models and their associated feature logic should execute in production, evaluating native Python, optimized libraries (e.g Polars), Spark, Ray, and specialized CPU/GPU formats (e.g., ONNX) against performance, scalability, operational complexity, and total cost of ownership.
Improve model-level efficiency: Identify opportunities to optimize model configurations, computational requirements, feature processing, and model representations while maintaining acceptable predictive quality and business outcomes.
Design for inference scalability: Investigate approaches to reducing duplicated computation across models, including shared feature computation and reusable intermediate representations, with the goal of improving inference efficiency as model complexity grows.
Integrate with rAVE: Define requirements, payload contracts, and integration patterns for productionized models running on rAVE or any real-time inference components. Collaborate with core runtime engineers to diagnose performance bottlenecks and evaluate improvements across model-to-runtime boundary, while maintaining clear ownership of model productionization versus inference-runtime implementation.
Build reusable engineering capabilities: Establish standards and tools for model packaging, dependency management, testing, versioning, reproducibility, and automated production delivery.
Drive end-to-end performance and cost improvements: Measure productionization lead time, execution efficiency, and infrastructure costs, turning findings into concrete engineering improvements and architectural decisions.
Provide technical leadership: Guide implementation decisions, mentor engineers, and align Data Science, ML Platform, feature infrastructure, and inference-runtime teams around clear technical interfaces and ownership boundaries.
About you as a Lead ML Systems Engineer
- Strong software engineering background, with deep Python expertise and experience designing reusable frameworks or platforms for production ML or data-processing workloads
- Proven experience designing and operating production ML systems, including model lifecycle management, packaging, validation, deployment, and monitoring
- Strong understanding of computational performance, including profiling, memory management, parallelism, data-processing overhead, and resource utilization
- Experience making architectural trade-offs across data-processing systems, data access patterns,execution environments, distributed processing, model-serving systems, and cloud infrastructure
- Ability to evaluate the economics of ML systems, balancing model complexity, inference throughput, latency requirements, infrastructure utilization, and total cost of ownership
- Experience designing systems that integrate multiple models, shared features, or common computational components
- Familiarity with Rust or experience working across language boundaries, with the ability to establish strict contracts and guide architectural integration with a Rust-based inference runtime(s)
- Experience with cloud infrastructure, automated delivery pipelines, production observability, and reliability engineering
- Strong technical leadership, including the ability to establish architecture, influence cross-functional teams, and turn ambiguous problems into an executable engineering roadmap
Technologies and Experience Core technologies
- Python and the data-processing ecosystem, including NumPy, pandas, and PyArrow
- MLflow and ML lifecycle-management practices.
- Databricks, Spark/PySpark, and production data-processing pipelines.
- Cloud infrastructure, preferably GCP (GCS, GKE),, including containerization and CI/CD and workflow orchestration (Airflow)
- Production ML systems, model-serving interfaces, performance profiling, and observability
Additional experience that is a plus
- Rust and integration with high-performance inference runtimes.
- ONNX, ONNX Runtime, nvForest, GPU training (PyTorch), GPU inference, or neural-network serving
- SQL and experience working with OLAP and/or OLTP systems.
- Understanding of analytical and transactional data architectures, data access patterns, and their implications for feature extraction, curation, and online serving
- Bigtable, Aerospike, or similar low-latency data stores.
- Online feature stores, real-time feature pipelines, and distributed data-serving architectures
- Experience with advertising technology, real-time bidding, or other high-throughput, latency-sensitive systems
- Experience optimizing infrastructure costs for compute-intensive or large-scale ML workloads
What Success Looks Like
A clear, technically validated architecture for moving models from experimentation into production across appropriate execution environments
A reusable productionization framework that reduces duplicated engineering effort and enables Data Scientists to deliver models more independently
Measurable improvements in productionization lead time, data and feature pipeline efficiency, inference performance, resource utilization, and total cost of ownership
A validated architectural strategy for scaling inference economically as model complexity grows, with measurable reductions in duplicated computation and improvements in cost per prediction or inference throughput where feasible
A coherent approach to integrating CPU and GPU inference capabilities (e.g., ONNX, nvForest) with rAVE and the wider ML Platform through predictable and strict engineering contracts
Consistent engineering standards for model validation, packaging, deployment, observability, and production readiness
Effective technical alignment across Data Science, ML Platform, feature infrastructure, and inference-runtime teams
Lead execution-platform evaluation: Define an evidence-based roadmap per workload type, including GPU training for neural-network models
Data-backed platform decisions per workload, and a clear path from notebook-based to IDE-native development
DT (NASDAQ: APPS) is a mobile growth platform that powers superior experiences and outcomes for advertisers, app developers, carriers, OEMs, and publishers worldwide.
Our technology connects brands and apps with consumers across devices, apps, and mobile environments — helping partners drive awareness, acquisition, engagement, and monetization through intelligent, integrated experiences.
Headquartered in Austin, Texas, DT operates globally across North America, EMEA, APAC, and LATAM.
DT is an equal opportunity employer committed to fostering an inclusive and diverse workplace. We welcome people of different backgrounds, experiences, abilities, and perspectives, and believe teams with different points of view build better products, ideas, and experiences.
Job description provided by Digitalturbine.
View the original posting