Shreya Shankar

Who is Shreya Shankar?

‍Shreya Shankar is a UC Berkeley computer science researcher known for work on LLM evaluation, data quality, and the day-to-day workflows of AI engineers. Her research focuses on building practical systems for AI-powered data processing and observability. (sh-reya.com)

Background and career

‍Shreya Shankar is a researcher in computer science with a focus on databases and human-AI interaction. On her personal site, she says she is completing her PhD in EECS at UC Berkeley in the Data Systems and Foundations group, advised by Aditya Parameswaran, and that she will join Carnegie Mellon University as an assistant professor in 2027. (sh-reya.com)

‍Her work sits at the intersection of systems, interfaces, and real-world AI use. She describes her goal as building the full stack for data-intensive knowledge work, and her current projects include efficient AI-powered data processing, AI agent observability, and long-running human-AI collaboration. (sh-reya.com)

‍Key facts about Shreya Shankar include:

1. Current role: UC Berkeley researcher in computer science, with an upcoming faculty role at Carnegie Mellon University in 2027. (sh-reya.com)
2. Research focus: LLM evaluation, data quality, AI-powered data processing, and AI agent observability. (sh-reya.com)
3. Training: PhD work in UC Berkeley EECS, in the Data Systems and Foundations group. (sh-reya.com)
4. Notable systems: DocETL and DocWrangler, built for scalable LLM-powered data processing. (sh-reya.com)
5. Public-facing work: papers, a course, and a book on evaluating LLM-powered applications. (sh-reya.com)

Notable contributions

‍

1. EvalGen: A mixed-initiative system for aligning LLM-generated evaluation functions with human requirements. (arxiv.org)
2. DocETL: An open-source system for declarative, LLM-powered document processing that uses rewrites and evaluation to improve accuracy. (arxiv.org)
3. SPADE: Work on synthesizing data quality assertions for large language model pipelines. (sh-reya.com)
4. PromptEvals: A dataset of assertions and guardrails for custom production LLM pipelines. (sh-reya.com)
5. AI-powered data workflow research: Broader work on making unstructured data processing more reliable, usable, and cheaper in practice. (sh-reya.com)

Why they matter in AI today

‍

1. They connect evals to real work: Their research treats evaluation as part of production workflows, not just benchmark scoring. (arxiv.org)
2. They focus on data quality: Many LLM failures come from messy inputs and vague tasks, and her work targets those failure modes directly. (sh-reya.com)
3. They bridge AI and systems: The research translates database ideas like rewrites and optimization into LLM pipelines. (sh-reya.com)
4. They reflect how engineers actually build: The work emphasizes reusable tools, iterative debugging, and human-in-the-loop design. (sh-reya.com)
5. They are highly applied: The projects are built with real users and real deployments, which makes the lessons practical for AI teams. (sh-reya.com)

Where to follow their work

‍Her personal website is the best starting point for publications, projects, and research updates. She also links to her Twitter, GitHub, Google Scholar, and CV from that page. (sh-reya.com)

‍For a deeper look at her current thinking, her research statement lays out her goals around AI-powered data systems, evaluation, and interface design. (sh-reya.com)

‍She is also associated with Berkeley research pages that list her as a PhD researcher in the Data Systems and Foundations group. (besi.berkeley.edu)

How PromptLayer connects with Shreya Shankar's work

‍Shreya Shankar's research is a strong fit for teams that need to manage prompts, track evaluations, and debug real LLM workflows. PromptLayer helps product and engineering teams bring that same discipline into daily development, with prompt management, observability, and iteration around what actually works in production.

Ready to try it yourself? Sign up for PromptLayer and start managing your prompts in minutes.