Depending on your experience, you will focus on one or more of the following areas:
- POI Search & NLP Infrastructure
Own query understanding, content understanding, relevance, and POI retrieval, ranking, and result presentation. Leverage LLMs to understand multilingual expressions of places, categories, budgets, and search intents, and extract structured information from videos and location data to power search. Combine semantic retrieval, geolocation signals, and click/conversion prediction to improve the accuracy and relevance of restaurant, hotel, and other search results.
Apply these capabilities to the online search pipeline through LLM-based labeling, training data development, model fine-tuning, and distillation, and explore approaches such as generative retrieval.
- Video Ranking & Personalization
Own retrieval, ranking, and personalization modeling for local services video search. Use LLMs to interpret the places, services, and consumption experiences featured in videos, assess their relevance and information value to user queries, and feed these insights into training data and ranking models. Optimize multiple objectives — including clicks, content consumption, and transaction conversion — using user interests, location, and behavioral sequences, so users can more easily find videos that are relevant to their immediate needs and rich in useful information.
- AI Search & Local Services Agent
Build agentic search experiences for restaurant, hotel, and travel needs. Combine LLMs, retrieval-augmented generation (RAG), and multimodal understanding to organize evidence-based answers from videos, location data, and user experiences, helping users compare options. Explore multi-turn intent clarification, context and preference memory, and tool use, enabling users to progressively refine their searches by location, budget, party size, and more. Contribute to model training, effect evaluation, and the rollout of search products.
- Algorithm Delivery & Optimization
Drive data infrastructure, model training, offline evaluation, online serving, and A/B experimentation. Continuously improve search quality and business outcomes based on user feedback, while balancing model performance, response latency, and computational cost.