Senior Machine Learning Engineer

I've spent 7+ years architecting AI/ML systems that bridge the gap between cutting-edge research and real-world impact. My journey started at Stanford, where I worked with pioneers like Fei-Fei Li on generative AI and visual understanding—research that grounded me in the fundamentals of representation learning and multi-modal AI. But what truly sets me apart is my unwavering commitment to turning innovation into scale. At DiDi, I didn't just experiment—I built reinforcement learning systems for autonomous vehicles processing 500GB+ datasets, architecting distributed training infrastructure that reduced training time by 40%. At TikTok, I scaled that vision further: designing demand forecasting systems that improved inventory accuracy by 25%, building conversational AI shopping assistants with RAG-based retrieval, and implementing graph neural networks that drove a 15% increase in customer lifetime value. Now at DoorDash, I'm leading ML ranking systems for ads and promotions, integrating LLMs and multi-modal embeddings into real-time inference platforms.

What makes me unique:

1. Research-to-Production Pipeline: Most engineers specialize in one or the other. I've published research on generative AI and scene understanding while simultaneously shipping production systems at scale across e-commerce, autonomous vehicles, and supply chain optimization.
2. Full-Stack AI Systems: I don't just train models—I design end-to-end platforms. From data pipelines processing 500M+ events daily (PySpark, Kafka, AWS), to vector databases and RAG systems, to frontend dashboards (React, FastAPI) that surface insights in real-time. I think in systems, not isolated algorithms.
3. Multi-Domain Expertise: Reinforcement learning for autonomous driving, demand forecasting with Transformers, graph neural networks for user segmentation, document intelligence with OCR and LLMs, conversational AI with voice interaction—I've solved hard problems across diverse domains. This adaptability means I can quickly synthesize insights and solutions across complex problem spaces.
4. Impact at Every Scale: Whether optimizing for 25% inventory improvements, 30% stockout reduction, 95% trajectory prediction accuracy, or 15% CLV growth—I measure success by business outcomes. I'm equally comfortable with cutting-edge LLM architectures and pragmatic solutions that actually move metrics.
5. Leadership in Emerging Tech: I didn't just adopt LLMs when they became trendy—I've architected RAG systems, multi-agent AI using LangGraph, document intelligence pipelines with Azure OpenAI, and conversational AI systems from first principles. I lead, not follow, in emerging technologies.