About Us
Mintegral is a leading programmatic and interactive mobile advertising platform, starting from the APAC region and radiating out globally. Powered by advanced AI technology, we provide global advertisers and developers with innovative, comprehensive experiences. With our efficient mobile marketing and monetization solutions, we help our clients exceed their marketing goals.
As Mobvista’s self-developed programmatic platform, since Launched in 2015, Mintegral has quickly grown to become one of the largest mobile advertising platform in Asia. We offer a full stack of programmatic products and services including our Self-service Platform, DSP, SSP, Ad Exchange and DMP. We have also created the Mindworks Creative Studio, which offers publishers and brands cutting-edge creative solutions, from traditional creative right through to the latest interactive ad formats. For more information, please visit our website: https://www.mintegral.com/en/
About the Role
We are looking for a systems expert to architect the next generation of our advertising engine. This role sits at the intersection of High-Performance Computing and Large-Scale Machine Learning. You will not just build infrastructure; you will define how our models learn and serve at a global scale, ensuring our systems are as intelligent as they are efficient.
Why This Role?
You will have the autonomy to rethink our core stack. Rather than just managing existing pipelines, you will be empowered to introduce new compilation techniques and system architectures that redefine our competitive advantage in the advertising market.
Responsibilities
- High-Performance Inference & Compilation: Architect and optimize real-time inference engines for ultra-large-scale models. Leverage AOT compilation, kernel optimization, and hardware-accelerated runtimes to push the limits of latency and throughput in heterogeneous environments (CPU/GPU/ASIC).
- Scalable Training Infrastructure: Design and evolve distributed training systems capable of handling multi-terabyte embedding tables and sparse parameters. Solve the complex sharding and synchronization challenges inherent in global-scale recommendation models.
- Unified Feature & Serving Architecture: Bridge the gap between feature engineering and model serving. Build high-efficiency systems for real-time feature generation and transformation, ensuring strict consistency between offline training and online execution.
- System-Business Co-design: Collaborate with Algorithm and Product teams to translate business needs—such as model freshness and ranking precision—into system-level capabilities. Optimize the end-to-end "data-to-dollars" pipeline for the advertising business.
Required Qualifications
- Systems Mastery: 5+ years of experience in AI Infrastructure, Distributed Systems, or High-Performance Computing. Proficiency in C++ and Python is essential.
- Deep Learning Internals: Strong understanding of ML framework internals (e.g., PyTorch, JIT, or specialized compilers like MLIR/TVM) and experience optimizing for GPU architectures (CUDA/Triton).
- Ads/Recommendation Context: Experience with the unique challenges of Ad Tech, including sparse feature processing, low-latency ranking, and high-QPS retrieval systems.
- Architectural Vision: Proven ability to lead the migration from legacy architectures to modern, hardware-efficient frameworks.
- Business Curiosity: A genuine interest in how technical performance (latency, efficiency) directly drives business ROI and advertising auction dynamics.
