Newsletter
Innovator profile
JL

Jerry Liu

Co-founder and CEO, LlamaIndex

AI agentsRetrieval-augmented generationDocument AIOpen source

Princeton, Quora and Uber ATG

Jerry Liu graduated summa cum laude from Princeton in 2017 with a BSE in computer science and a certificate in finance[1]. He then worked as a machine learning engineer at Quora on its feed ranking system[1].

Liu moved to Uber's Advanced Technologies Group, joining as an AI resident in 2018 and becoming a research scientist in 2019[2]. There his research covered learning-based compression and joint prediction and planning for autonomous driving[1]. His papers from that period include DSIC: Deep Stereo Image Compression at ICCV 2019 and Conditional Entropy Coding for Efficient Video Compression at ECCV 2020, both with Raquel Urtasun[1]. He is the first author of Deep Structured Reactive Planning, with Wenyuan Zeng, Raquel Urtasun and Ersin Yumer, presented at ICRA 2021[3].

He then became a machine learning engineering manager at Robust Intelligence, where he led work on model monitoring and the company's AI Firewall[1][2].

From GPT Index to LlamaIndex

In November 2022 Liu created GPT Index, a set of data structures that let GPT-3 traverse external data[1][2]. He told TechCrunch that it started as a toy open-source project because he wanted to understand how large language models could be used on proprietary data outside their training set[4]. The project grew into a framework with connectors, indexing, retrieval and answer-synthesis components and became LlamaIndex[2].

In 2023 Liu and Simon Suo, also a former Uber research scientist, founded LlamaIndex as a company[4]. They announced an $8.5 million seed round led by Greylock[2]. In March 2025 the company raised a $19 million Series A led by Norwest Venture Partners, with Greylock taking part, bringing its total funding to $27.5 million[4][5]. TechCrunch reported that it had a 20-person team at that time and that Salesforce, KPMG and Carlyle used its software[4].

Document parsing and agents

In February 2024 Liu announced LlamaParse, a managed service for parsing, ingesting and retrieving data from complex documents with embedded tables and figures[6]. The company made LlamaParse generally available with its Series A in March 2025, offering it as a hosted service or an on-premises installation[5]. TechCrunch described the company's cloud service as a way for customers to build agents that work with unstructured data in many formats[4].

Liu's talks focus on how models lose the structure of documents[2]. He has used the example of a train timetable, where a parser can keep every word and number but lose which time belongs to which column[2]. He recommends measuring retrieval and generation separately and testing parsing, chunk size, hybrid search and metadata filters before adding more costly agent designs[2]. He also argues for two-pass processing, in which a cheap first pass makes a collection searchable and a deeper visual pass examines only the relevant pages[2].

In March 2026 LlamaIndex open-sourced LiteParse, a command-line tool and TypeScript library that parses layout-aware text from PDFs, Office documents and images entirely on a local machine[7]. On 9 September 2026 Liu announced a LlamaParse connector for ChatGPT, following an earlier connector for Claude[2]. His AI Engineer speaker page describes the company's focus as shifting towards LlamaParse, LiteParse and open parsing and extraction benchmarks, with the original framework still available[2].

Known for

Sources

  1. Jerry Liu's personal website — checked 6 Nov 2026
  2. AI Engineer: speaker page, Jerry Liu — checked 6 Nov 2026
  3. arXiv: Deep Structured Reactive Planning, January 2021 — checked 6 Nov 2026
  4. TechCrunch: LlamaIndex launches a cloud service for building unstructured data agents, March 2025 — checked 6 Nov 2026
  5. LlamaIndex blog: Announcing our Series A and LlamaParse general availability, March 2025 — checked 6 Nov 2026
  6. LlamaIndex blog (Jerry Liu): Introducing LlamaParse, February 2024 — checked 6 Nov 2026
  7. LlamaIndex blog: LiteParse, local document parsing for AI agents, March 2026 — checked 6 Nov 2026
Editorial profile compiled from public sources. Not an endorsement, and listing does not imply participation in this site. To correct or remove this profile, contact us.