subash luitel

I’m a Senior Machine Learning Engineer with 10+ years developing large-scale ML systems across advertising, computer vision, and real-time inference for teams at Google, Pinterest, and Lockheed Martin. I focus on enhancing model performance, minimizing latency, and scaling ML infrastructure for production environments reaching millions of users.

At Lockheed Martin, I developed and deployed computer vision models for mission-critical autonomous and sensor-processing systems, improving detection accuracy and operator situational awareness. I built real-time perception and tracking pipelines, reducing inference latency by ~30–40% and improving robustness under noisy conditions, while leading end-to-end ML lifecycle work from data to training to deployment. I’ve also integrated ML into embedded and simulation systems using Python and C++, and strengthened delivery through MLOps standardization.

Earlier, at Pinterest, I built ads ranking and sequence modeling systems for long-term user behavior, reducing serving latency by ~20% while maintaining performance. I designed real-time pipelines with Kafka and Flink for low-latency recommendation workflows, and led video ML infrastructure and high-throughput inference services for quality scoring. At Google, I engineered high-throughput C++/Java backend services for search and ads serving, built distributed feature pipelines using MapReduce and Dataflow-style systems, and improved reliability through monitoring, alert tuning, and incident response.