• Drove technical alignment in redefining ranking systems across Ads, Content and Growth core products.

• Led technical design of online learning recommendation system (data, training, inference), increased model parameter freshness to O(min) improving topline metrics.

• Defined and executed a multi-year ML Platform transformation enabling terabyte-scale recommender model training and inference.

• Pioneered dynamic embeddings eliminating existing bottlenecks, introduced inference-time embeddings store scaling capability beyond single-host memory.

• Built custom kernels for remote embeddings lookups, optimized runtime execution by pooling operations for latency reduction.

• Partnered with ML product teams, scaled core product recommender models x10 up to 1TB improving user engagement.

• Designed and scaled foundational ML infrastructure, enabling reliable, low-latency production workloads at internet scale.

• Engineered a distributed LSM-backed online feature store, onboarded core use cases exceeding 100M keys/sec throughput and 16Tbps network egress.

• Established the retrieval platform team, drove adoption of embedding-based retrieval pipelines across the company.

• Founding member of the AI Platform, built and scaled the Inference Platform exceeding 1K concurrent experiments, 300T predictions/day, 1.5M CPU cores, and 50K GPUs.

• Optimized high-performance compute workloads, inference runtimes and computation graphs reducing latency and improving system efficiency.

Designed and deployed dynamic pricing system for private jet charters, optimizing revenue per flight hour and inventory utilization across variable demand, routing and availability constraints.