Machine Learning Engineer

Who you are

• Master's degree in machine learning or a related field required; a PhD is a plus
• At least 3 years of experience as a Machine Learning Engineer, building state-of-the-art machine learning/deep learning systems at extreme scale
• Expertise in large-scale distributed data systems, including high-performance relational and key-value stores, orchestrators like Airflow, and data transformations in Spark
• Expertise in Python; familiarity with Java (especially in high-performance serving systems) is a plus
• Expertise in PyTorch or a similar ecosystem
• Experience with accelerator-based inference systems such as Triton, and optimization backends like ONNX and TensorRT
• Experience with cloud platforms and container orchestration using Kubernetes
• Knowledge of deep recommender systems, including two-tower and attention-based architectures, is a plus
• Knowledge of reinforcement learning systems — including RL approaches, reward signals, and post-training feedback loops — is a plus
• Understanding of the ads ecosystem (OpenRTB, SSPs, DSPs, ad exchanges, MMPs) is a plus

What the job involves

• We are looking for a Machine Learning Engineer to develop and optimize our ad science models, open-source LLMs, and the underlying infrastructure that powers them — with a constant eye toward performance and cost
• AI and ML sit at the core of InMobi's business, and our science teams are tackling some of the field's most interesting problems: creative optimization, user personalization, targeting, yield management, traffic shaping, and more
• We're moving fast and deep into architectural research, generative AI adoption, compute scaling on the latest accelerators (GPUs/TPUs), and advances in reinforcement learning
• You'll join a team of experienced researchers and engineers who have been pushing the boundaries of ad science research for over a decade, working closely with data and platform engineering to build solutions that scale and serve the needs of every science team across the organization
• This is a chance to shape the infrastructure and models behind ad intelligence operating at truly massive scale
• Build and support training pipelines and model implementations that maximize experimentation velocity
• Adapt open-source LLM infrastructure like DeepSpeed and OpenRLHF to meet InMobi-specific post-training needs
• Optimize online and batch inference for low latency and cost efficiency
• Build monitoring and evaluation solutions that keep our infrastructure reliable and our models behaving as expected
• Leverage the breadth of data features across our product portfolio to strengthen our data pipelines and unlock new features for training and inference
• Explore new platforms and ecosystems (e.g., JAX on TPUs) to help diversify our compute
• Collaborate closely with Applied Scientists on active research, contributing directly to experiment design and modeling