AI Software Engineer Apple
I progressed from large-scale production software and distributed data systems into machine-learning infrastructure, search and ranking, on-device AI, foundation models, and MCP-enabled agentic applications. I led system architecture, production reliability, model evaluation, inference optimization, privacy-aware design, and cross-functional engineering initiatives.
I designed distributed ML workflows for data ingestion, validation, transformation, feature generation, training, evaluation, artifact management, reproducibility, dataset lineage, model versioning, release qualification, staged deployment, monitoring, and rollback procedures. I supported production workflows for an approximately 3-billion-parameter on-device foundation model serving 15 languages and developed multilingual evaluation workflows using 1,000 native-speaker prompts per supported language.
I built personalized search and intelligent ranking workflows using textual relevance, behavioral signals, personalization, semantic similarity, contrastive-learning embeddings, XGBoost, candidate retrieval, reranking, and online scoring. I developed an LLM-assisted relevance-labeling workflow that generated millions of labels and contributed to a worldwide A/B test producing a statistically significant 0.24% conversion-rate increase.
I designed on-device inference and model-optimization workflows across CPU, GPU, and Apple Neural Engine, including quantization, compression, profiling, device benchmarks, and compatibility testing. I also architected MCP-based agentic AI systems with secure tool discovery and execution, RAG, multi-step orchestration, OAuth authorization, audit logging, prompt-injection defenses, model routing, and observability.