Peilu Zhang

Roadblock title:

Reducing manual QA processes by 50%: Harnessing AI for faster and smarter data assessment

Time:

Thursday - 1:00 PM (5th floor - Manchester E)

Abstract:

As a company specializing in data, data quality is a critical metric for our product’s performance. Traditionally, data companies will invest substantial resources into manually assessing data quality to ensure accurate and high quality data, which can be a time consuming process and slow down delivery of features. At Spade, we employ both manual and programmatic techniques to perform data QA, including advanced capabilities of ChatGPT to streamline data review workflows. We’ll explore how we leverage LLMs to automate our QA process and turnaround results in a fraction of the time of the old process. We’ll delve into our approach, as well as how we enhanced the performance with prompt engineering and enabling LLM to undertake more complex tasks informed by the outcomes of human QA.

Bio:

Peilu started her career as a financial engineer, and made the transition to tech over 6 years ago. Since then, she has been using her Machine Learning skills to solve customer pain points at companies big and small in the FinTech space. She helped build out the Fee Negotiation AI Agent at Harvest, which contributed to a successful acquisition by Acorns, and is now building out the ML capabilities at Spade, a real-time merchant intelligence API. Her product prowess and empathy for the customer motivates her to build solutions that are pragmatic and resilient. Her latest work focuses on making unstructured data structured and useful, using a variety of technologies and methods including Graph Data Structures and LLMs.