About
co-founder/CEO of LlamaIndex
Building the world's best agentic document…
Building the world's best agentic document…
Activity
50K followers
-
Jerry Liu reposted thisLove to see records spanning pages jump from 85.5% to 96.5%. That's where we most often see customers getting stuck with other tools. The advanced citations also make it very easy to trust the output. Great work from the team! Excited to show this to our partners!!
-
Jerry Liu reposted thissince i've joined LlamaIndex, one of the indicators that has been the greatest confirmation that i made the right decision was how fast this team moves. which is why i'm so hyped to share that we've officially launched extract v2.5! ⚡️ the team rebuilt the agent harness behind extract to go after crucial problems: lists that stop early, records that get chopped off at a page break, scanned permits where a reviewer's scribbled date ends up glued to the permit number. tldr: every tier now beats the tier above it from the last version for the same exact price if you're on another tool and have been meaning to see how it compares, we're giving $3,000 in credits to test the switch. give us your worst docs! :)
-
Jerry Liu reposted thisJerry Liu reposted thisI'm excited to announce our new Extract v2.5 release! This version is a significant jump, and at the cutting edge of document extraction capabilities. Looking forward to seeing what folks build https://lnkd.in/gGG9qxKa
-
Jerry Liu shared thisToday we’re introducing Extract v2.5 - a series of frontier agents tuned for document extraction. The agents (cost-effective, agentic, agentic plus) are tuned for value accuracy and grounding. Our extraction agents outperform Opus 5.5 and GPT-6 Sol while being 30%-4x cheaper. We’ve made massive improvements on complex extraction over * long lists (86.1% -> 95.5% on our agentic tier) * records spanning pages (85.5% -> 96.5%) * scanned forms (90.9% -> 95.7%) We’ve also launched the following features: ✅ Advanced citations: we locate bounding boxes for all supporting values for an inferred field, even if there's not an exact match. ✅ Structural Reasoning: we tailor document extraction algorithms depending on the type, layout, and information Our extraction agents are SOTA in price-performance on ExtractBench, across a wide range of cost points. We are the best tool for document extraction across documents of any complexity. Blog: https://lnkd.in/gPfUnNna All of these are available on LlamaParse: https://lnkd.in/g9Wpqn7w . Come check it out!
-
Jerry Liu reposted thisJerry Liu reposted thisThe race to superintelligence is an infrastructure race 🏁 At Fully Connected 2026, #theCUBE takes you inside CoreWeave on day two as it shows how developer platforms and tools simplify deploying and managing AI workloads, and how the push to cut time to first token is reshaping infrastructure. We also look at how liquid cooling and data center design are changing to support high-density AI systems. Hosted by David Vellante and John Furrier, we'll unpack the role of neocloud providers in expanding access to specialized AI infrastructure, customer examples of large-scale training, inference and agentic applications, and the technical and economic realities behind the pursuit of superintelligence. Hear directly from: 👉 CoreWeave's Chen Goldberg Lukas Biewald, Jim Higgins and Susanne Seitinger, PhD 👉 NVIDIA's Ian Buck 👉 Cognition's Silas Alberti 👉 VAST Data's Alon Horev 👉 LlamaIndex's Jerry Liu 👉 Willow's Lawrence Liu 👉 Liquid AI's Jeffrey Li 👉 BattleBots' Robert Bahr 👉 CrowdStrike's Bartley Richardson 📺 Watch live, only on theCUBE: https://lnkd.in/dnQ5qywH #FullyConnected #CoreWeave #AgenticAI #DeveloperExperience
-
Jerry Liu shared thisSystem one models are extremely useful for a variety of document tasks that require fast decisions: * orientation detection * language detection * classification * splitting We benchmarked Jev with other OSS models (some general classifiers, some document-specific) against a variety of these tasks. We measured accuracy, cost, and latency. Jev tops most of the benchmarks here across the 3 dimensions. We're excited to use it and learn from it to continue advancing the frontiers of doc understanding! YT: https://lnkd.in/grFjEFR3 Repo: https://lnkd.in/gghbQ6EB
-
Jerry Liu shared thisWe comprehensively benchmarked gpt-6.1 sol on document OCR tasks. compared to gpt-6 sol from a week ago (!), we see a sizable increase in table parsing capabilities and reading order, which is pleasantly surprising. gpt-6.1 sol has similar table parsing as gpt-6 astra. these are great results. frontier models are still typically an order of magnitude more expensive than cost-effective document parsing solutions, but this also means that there's a lot of alpha in improving capabilities at the "premium" end (>1c per page) of the spectrum too. check out ParseBench: https://www.parsebench.ai/ for a comprehensive set of document parsing capabilities, check out llamaparse: https://lnkd.in/g9Wpqn7w
-
Jerry Liu shared thisYesterday I hosted a fun dinner conversation with Vincent Sunn Chen from Snorkel AI on evals and RL environments. The "data and RL env" companies (like Snorkel) have seen massive growth in the past few years. There's been an explosion of interest in evals. At the same time, models are ripping through benchmarks with each new release. We talked about the evals everyone is defining, what evals are still left unsolved, what’s left up to frontier models vs. intelligence that you own, and more: - A big challenge for building RL environments is “fairness” - when the model fails on a given environment, can you attribute it to the input, harness, or reward model? - Building proper rewards is hard. Some tasks are not easily quantifiable. You also want to discourage reward hacking. At the same time, you don’t want to be too prescriptive with intermediate rewards. - Long horizon evals are still extremely hard, some business processes can take up to weeks or months before the final outcome - Most regulated industries still need human in the loop to guarantee ~100% accuracy, “80%” accuracy is not good enough - As models get more intelligent, there will be a barbell of boutique data vendors (e.g. any SMB) any scaled up data providers. - Models still exhibit “jagged intelligence” where they still fail on a long tail of edge cases. - There might always be opportunities to gather unique data for a given task and posttrain models for lower cost and higher accuracy. This marks #003 in our founder dinner series. What topic should we discuss next? Let us know your thoughts below! Preston Yadegar Faraz Siddiqi Akshaya Jagadeesh Ali Ebrahim Shahul Elavakkattil Shereef Robert Stewart Jithin James Abhijeet Shenoi Vincent Sunn Chen
-
Jerry Liu reposted thisJerry Liu reposted thisParsing forms still trips up the latest frontier VLMs. A form isn't text on a page. It's a set of fields, grouped into sections, each tied to a specific box. That's why forms need purpose-built parsing, not a bigger general model: ✅️ Detect every field, not just the obvious ones ✅️ Keep the hierarchy of sections and fields ✅️ Tie every value to the exact box it came from ✅️ Understand handwriting and checkmarks Our latest blogpost breaks down where VLMs fail on real W-2s, 1040s, W-9s and scanned W-4s, and a custom cookbook for LlamaParse to handle them at a fraction of the cost 👇 https://lnkd.in/gzBPmtBa
Jerry Liu commented on a post
2w
congrats on the launch!
Jerry Liu commented on a post
2w
if you're here and don't want to use flies to parse your pdfs, you could always check out an actual OCR solution like llamaparse https://
cloud.llamaindex.ai
Jerry Liu replied to a comment
2w
i agree, it's a bit unfortunate 😔
-
Jerry Liu liked thisLove to see records spanning pages jump from 85.5% to 96.5%. That's where we most often see customers getting stuck with other tools. The advanced citations also make it very easy to trust the output. Great work from the team! Excited to show this to our partners!!
-
Jerry Liu liked thissince i've joined LlamaIndex, one of the indicators that has been the greatest confirmation that i made the right decision was how fast this team moves. which is why i'm so hyped to share that we've officially launched extract v2.5! ⚡️ the team rebuilt the agent harness behind extract to go after crucial problems: lists that stop early, records that get chopped off at a page break, scanned permits where a reviewer's scribbled date ends up glued to the permit number. tldr: every tier now beats the tier above it from the last version for the same exact price if you're on another tool and have been meaning to see how it compares, we're giving $3,000 in credits to test the switch. give us your worst docs! :)
-
Jerry Liu liked thisJerry Liu liked thisI'm excited to announce our new Extract v2.5 release! This version is a significant jump, and at the cutting edge of document extraction capabilities. Looking forward to seeing what folks build https://lnkd.in/gGG9qxKa
-
Jerry Liu liked thisJerry Liu liked thisI’m super excited to announce that Chronosphere is now Cortex XCOR - our next-generation, AI-native observability platform that provides automated root cause analysis and remediation. Chronosphere was the best placeholder name that Rob Skillington and I could think of on short notice as the company name choice was the main thing blocking investors from wiring us money early on. We thought we would change it later, but we never did - it grew on us and has served us well over the years. It became a huge part of our personal identities and was the title of that rewarding chapter of our lives. It is fitting that we have a new title and name for this next chapter: Cortex XCOR. However, it’s not just a new name / rebrand - it’s the biggest release of our platform ever. One where we reimagined the Observability experience and aimed to deliver the holy grail - a solution that automates root cause analysis and gives you the answer. You’ve probably seen a ton of solutions claiming this along with cool demos, but what impresses me most about what the team has been working on are the results we’re seeing in large production environments: investigations completed in under three minutes on average with a 75% success rate in identifying the root cause. For those who are still concerned about the economics of Observability, don’t worry - we haven’t abandoned our core principles. We’re still helping our customers optimize their data volumes by 89% on average, earning the top ranking Observability platform for cost efficiency and the trust of industry leaders like DoorDash, Affirm and OpenAI. Read more about XCOR in my blog post: https://lnkd.in/gsF7D6cZ
-
Jerry Liu liked thisJerry Liu liked thisI’m super excited to join Cognition as its Chief Revenue Officer. Going from the 1st sales rep at Snowflake to CRO at a $100b public company was the thrill of a lifetime. Never in my wildest imagination did I think I'd get the opportunity to jump on another rocket ship. Meeting Scott, Russell, and the entire team at Cognition got me so excited to do it again with world-class technology. Hearing how customers spoke about how Devin is transforming their business made me realize this is a once-in-a-lifetime opportunity. While I am not jumping on as the first sales rep this time, I can’t wait to dive in and help drive amazing customer outcomes while building a world-class sales team. As many of you know, Chad Peets was my co-pilot in building the sales team at Snowflake and my business partner when I joined RPT. There is no world where I could be the best version of myself without Chad. So I am thrilled that Chad and the RPT team will be working closely with me in our pursuit of building a generational company. LFG!!
Experience & Education
Publications
-
MuSCLE: Multi Sweep Compression of LiDAR using Deep Entropy Models
NeurIPS 2020
Sourav Biswas, Jerry Liu, Kelvin Wong, Shenlong Wang, Raquel Urtasun
-
DSIC: Deep Stereo Image Compression
ICCV 2019 (Oral)
See publicationJerry Liu, Shenlong Wang, Raquel Urtasun
-
Interactive 3D Modeling with a Generative Adversarial Network
International Conference on 3D Vision, 2017
Honors & Awards
-
Outstanding Senior Thesis Award
-
-
Shapiro Prize for Academic Excellence
Princeton University
Other similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content