Jinal Parikh


Jinal Parikh is a Software Engineer at Google and a speaker. She has been working on Fraud Detection for Maps, currently leading multiple LLM projects for content moderation and optimising processes. She has 3 years of ML engineering & Data Science experience in feature engineering, data analysis, deploying models for detecting monetary fraud on Maps. Having worked previously with Goldman Sachs and Morgan Stanley, she has now a total experience of 6 years in both the FinTech and Product space. She is also a Google WTM Scholar 2017, performing outreach activities to foster a local community of women in Tech. In her spare time, she does all things art!


Build low-latency, smarter LLMs with RAG


With the advent of LLMs in every industry, everyone has access to them. So your business’s superpower differentiation comes from feeding it with your data while being cost-effective- and that is what retrieval-augmented generation (RAG) enables us to do. In this talk, we will explore the architecture of RAG and how to build practical applications, such as chatbots and search enhancement. By walking through diverse example applications end-to-end, Jinal will demonstrate how to integrate external databases like vector databases and Neo4j. The talk will equip you with a reference architecture for RAG applications and insights into choosing the most suitable technique for your specific needs. You’ll also learn the answer to the classic question: RAG vs Fine-Tuning vs PEFT! Whether it’s RAG, prompt engineering, fine-tuning, or pre-training, you’ll learn how to discern the best approach for your application!