Shreya Shankar

Career

Shankar studied computer science at Stanford, completing a B.S. advised by Pat Hanrahan, and was a research intern at Google Brain from 2017 to 2019, working with Alex Kurakin and Ian Goodfellow on machine learning security and adversarial examples, and also worked as an ML engineer. Shankar began a PhD in EECS at UC Berkeley in 2021 with Aditya Parameswaran in the Data Systems & Foundations Group, with a research internship on Meta's AI Data Infrastructure team in 2022 and a visiting student researcher role at Google's Systems Research Group in 2025.⁶¹

Shankar's CV lists a Visiting Assistant Professor appointment in Carnegie Mellon University's Computer Science Department from July 2026, before a tenure-track Assistant Professor position there from July 2027, with a courtesy appointment in the Human-Computer Interaction Institute and an affiliation with the CMU Database Group. Shankar's group writes a blog called Full Stack Data Lab.⁶¹

Research

Shankar's research spans data management, AI-powered data systems and human-computer interaction. As first author of the 2024 DocETL paper, Shankar proposed agentic query rewriting and evaluation for complex document processing with language models, and the open-source DocETL stack grew out of that work at Berkeley's EPIC Data Lab.⁶⁴³

Follow-up work includes Task Cascades and "Cut Costs, Not Accuracy" at SIGMOD 2026, the latter recognized as Best of SIGMOD, and "RAG Without the Lag," an interactive debugging tool for retrieval-augmented generation pipelines that won a CHI 2026 Best Paper award. Shankar's group also released Data Agent Bench, a benchmark for data agents, and Quail, a query-aware inference engine for AI-SQL workloads.⁶¹

Teaching and writing

Since 2025 Shankar has co-taught AI Evals for Engineers and PMs, a professional course co-founded with Hamel Husain, and is co-writing a forthcoming O'Reilly book on the subject; the pair also publish an evals FAQ and other free materials. Shankar co-wrote the 2024 essay "What We've Learned From A Year of Building with LLMs" and has organized the DEEM workshop at SIGMOD from 2023 to 2025.⁶¹⁵

Work and contributions

• DocETL Open-source code

  Built DocETL during PhD research at UC Berkeley and is first author of "DocETL: Agentic Query Rewriting and Evaluation for Complex Document Processing" (2024). The MIT-licensed repository says DocETL was created at the EPIC Data Lab and Data Systems and Foundations group at UC Berkeley.¹⁴³

• What We've Learned From A Year of Building with LLMs Published research

  Co-author of the 2024 essay on lessons from building LLM products; the authors credit Shankar with bringing research and original results on evals and production practices.⁵

Timeline

1. 2017Began a research internship at Google Brain on machine learning security and adversarial examples.⁶
2. 2021Began a PhD in EECS at UC Berkeley with Aditya Parameswaran.⁶
3. 2022Interned with Meta's AI Data Infrastructure team on automatic data validation for ML pipelines.⁶
4. Oct 16, 2024Released the DocETL paper on agentic query rewriting for complex document processing as first author.⁴
5. 2025Began co-teaching the AI Evals for Engineers and PMs course with Hamel Husain.⁶
6. May 2026Published Task Cascades and "Cut Costs, Not Accuracy" in the SIGMOD 2026 proceedings.⁶
7. Jul 2026Began a Visiting Assistant Professor appointment in Carnegie Mellon University's Computer Science Department.⁶

Other documented roles

Roles the sources document that are not confirmed as current.

• PhD researcher in the Data Systems & Foundations Group, working with Aditya Parameswaran (2021–2026), University of California, Berkeley (EECS)²¹⁶

Honors and recognition

• Best of ACM SIGMOD · ACM SIGMOD, 2026⁶
• ACM CHI Best Paper · ACM CHI Conference on Human Factors in Computing Systems, 2026⁶
• ACM UIST Best Paper Honorable Mention · ACM Symposium on User Interface Software and Technology, 2025⁶
• NDSEG Fellowship · U.S. Department of Defense, 2022⁶

Where they work

• Carnegie Mellon University 24 people in the wiki

Related people

• Yulia TsvetkovAlso at Carnegie Mellon University
• Shinji WatanabeAlso at Carnegie Mellon University
• Karen LivescuAlso at Carnegie Mellon University
• Jiatong ShiAlso at Carnegie Mellon University
• Zico KolterAlso at Carnegie Mellon University
• Eugene YanEvaluation and benchmarks · Datasets and data
• Leo GaoEvaluation and benchmarks · Datasets and data
• Kyle LoDatasets and data · Evaluation and benchmarks
• Ruchir PuriDatasets and data · Evaluation and benchmarks
• Adina WilliamsEvaluation and benchmarks · Datasets and data