Tweets

traces matter!

A problem we see often: long documents with different pieces of repeating content. Example: a resume book with a cover page, a few pages about student curriculums, then back to back resumes

Build an intelligent resume processing agent that automatically extracts structured data from repeating content using LlamaSplit to identify where each individual content starts and ends and LlamaExtract to extract structured data:

📄 Upload PDF resume books to LlamaCloud and automatically categorize pages using LlamaSplit to separate individual resumes from curriculum and cover pages
🤖 Extract structured information from each resume using LlamaExtract with custom schemas to capture names, contact info, education, work experience, and skills etc (your choice)
⚡ Orchestrate the entire process with LlamaAgent Workflows
🔍 Process real resume data with confidence scores and structured output ready for filtering, searching, and candidate matching systems

The tutorial uses an NYU Resume Book as an example and shows both individual API calls and a complete automated workflow implementation.

Check out the full tutoriadevelopers.llamaindex.ai/python/cloud/l…Hq8

Insanely exciting to see AI in healthcare starting to WORK.

Researchers in China found an AI tool was better than radiologists at identifying pancreatic cancer.

Meanwhile at Stanford, a new foundation model can predict disease risk from a single night of sleep data 🤯

Enterprise knowledge has never been limited to text.

With the capabilities of Multimodal Embedding and Rerank, Dify now supports retrieving and reasoning over text and images together in a unified semantic space. This means diagrams, screenshots, and product photos can participate directly in RAG workflows as actionable context for vision-enabled LLMs.

The guide walks through the technical design, the Dify embedding–rerank synergy in multimodal RAG, and a practical workflow example.

Explore how it works: dify.ai/blog/multimoda…

#Dify #KnowledgeBase #RAG #VisionLLM

👗𝐒𝐧𝐚𝐩 𝐚 𝐜𝐥𝐨𝐭𝐡𝐢𝐧𝐠 𝐩𝐡𝐨𝐭𝐨 → 𝐠𝐞𝐭 𝐬𝐢𝐦𝐢𝐥𝐚𝐫 𝐢𝐭𝐞𝐦𝐬 𝐢𝐧 𝐦𝐢𝐥𝐥𝐢𝐬𝐞𝐜𝐨𝐧𝐝𝐬. @mozatnigeria 𝐛𝐮𝐢𝐥𝐭 𝐒𝐭𝐲𝐥𝐞𝐩𝐞𝐝𝐢𝐚 𝐟𝐚𝐬𝐡𝐢𝐨𝐧 𝐚𝐩𝐩 𝐨𝐧 𝐌𝐢𝐥𝐯𝐮𝐬 𝐭𝐨 𝐩𝐨𝐰𝐞𝐫 𝐡𝐲𝐩𝐞𝐫-𝐩𝐞𝐫𝐬𝐨𝐧𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐫𝐞𝐜𝐨𝐦𝐦𝐞𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐬 𝐚𝐭 𝐬𝐜𝐚𝐥𝐞.

Delivering this experience consistently is not trivial.
As Stylepedia’s image library grew rapidly, the team faced a dual challenge: 𝐬𝐜𝐚𝐥𝐞 𝐚𝐧𝐝 𝐬𝐩𝐞𝐞𝐝.

Stylepedia needed to efficiently 𝐬𝐭𝐨𝐫𝐞, 𝐫𝐞𝐜𝐨𝐠𝐧𝐢𝐳𝐞, 𝐜𝐨𝐦𝐩𝐫𝐞𝐡𝐞𝐧𝐝, 𝐚𝐧𝐝 𝐜𝐨𝐦𝐩𝐚𝐫𝐞 𝐢𝐦𝐚𝐠𝐞𝐬 of clothing that vary widely in resolution, size, and viewing angle—while still returning 𝐢𝐦𝐚𝐠𝐞 𝐬𝐢𝐦𝐢𝐥𝐚𝐫𝐢𝐭𝐲 𝐫𝐞𝐬𝐮𝐥𝐭𝐬 𝐰𝐢𝐭𝐡𝐢𝐧 𝐦𝐢𝐥𝐥𝐢𝐬𝐞𝐜𝐨𝐧𝐝𝐬.

To solve this, Mozat adopted Milvus as the vector similarity search engine at the core of Stylepedia’s image search system.
🚀𝐊𝐞𝐲 𝐑𝐞𝐬𝐮𝐥𝐭𝐬
• 𝐓𝐞𝐧𝐬 𝐨𝐟 𝐦𝐢𝐥𝐥𝐢𝐨𝐧𝐬 𝐨𝐟 𝐯𝐞𝐜𝐭𝐨𝐫 𝐞𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠𝐬 handled with fast scaling capability
• 𝐌𝐢𝐥𝐥𝐢𝐬𝐞𝐜𝐨𝐧𝐝-𝐥𝐞𝐯𝐞𝐥 𝐢𝐦𝐚𝐠𝐞 𝐬𝐢𝐦𝐢𝐥𝐚𝐫𝐢𝐭𝐲 𝐬𝐞𝐚𝐫�zilliz.com/customers/moza…𝐞𝐱𝐩𝐚𝐧𝐬𝐢𝐨𝐧 as the image library continues to grow
• 𝐇𝐲𝐩𝐞𝐫-𝐩𝐞𝐫𝐬𝐨𝐧𝐚𝐥𝐢𝐳𝐞𝐝 𝐟𝐚𝐬𝐡𝐢𝐨𝐧 𝐝𝐢𝐬𝐜𝐨𝐯𝐞𝐫𝐲 powered by visual similarity

With Milvus, Stylepedia evolved from a digital closet into an intelligent fashion companion — 𝐛𝐫𝐢𝐧𝐠𝐢𝐧𝐠 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫 𝐡𝐮𝐦𝐚𝐧 𝐜𝐫𝐞𝐚𝐭𝐢𝐯𝐢𝐭𝐲 𝐚𝐧𝐝 𝐦𝐚𝐜𝐡𝐢𝐧𝐞 𝐢𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞.

🔗 Full story: https://t.co/1Mwa3cQ4KO
———
👉 Follow @milvusio for real-world vector database use cases powering AI at scale.
#VectorDatabase #BuildwithMilvus

AI agents don’t just hallucinate—they suffer from 𝐂𝐨𝐧𝐭𝐞𝐱𝐭 𝐑𝐨𝐭. 🧠📉
As context windows grow, retrieval accuracy can drop by up to 30%. 𝐌𝐨𝐫𝐞 𝐭𝐨𝐤𝐞𝐧𝐬 𝐨𝐟𝐭𝐞𝐧 𝐥𝐞𝐚𝐝 𝐭𝐨 𝐦𝐨𝐫𝐞 𝐧𝐨𝐢𝐬𝐞, 𝐧𝐨𝐭 𝐦𝐨𝐫𝐞 𝐢𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞.

Here’s how to fix it with Context Engineering:
✅ 𝐉𝐮𝐬𝐭-𝐢𝐧-𝐓𝐢𝐦𝐞 (𝐉𝐈𝐓) 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥: Don’t overstuff the prompt. Fetch data only when the agent triggers a specific tool.
✅ 𝐇𝐲𝐛𝐫𝐢𝐝 𝐒𝐞𝐚𝐫𝐜𝐡: Use Milvus for stable, long-term knowledge and Agent Tools for real-time logs/APIs.
✅ 𝐒𝐮𝐦𝐦𝐚𝐫𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐨𝐩𝐬: Don't pass raw history. Summarize earlier steps to keep the "working memory" lean.milvus.io/blog/keeping-a…�𝐮𝐥𝐞: A small, accurate prompt beats a massive, noisy one every time.
Read the full deep dive here: https://t.co/BIDA9SFVC9
———
👉 Follow @milvusio, for everything related to unstructured data!

🌕 @gru_space is building durable space habitats so humans can one day live on the Moon and Mars.

Its first missions will mine lunar regolith to construct a long-term pressurized habitat on the Moon for commercial space tourism — a hotel on the Moon.

Congrats on the launch @skyler_chan_!
ycombinator.com/launches/P9g-g…U

Stoke Space is racing to build the world's first fully reusable rockets that can launch, survive reentry, and fly again almost right away.

In this episode of Hard Tech, @aaron_epstein sits down with @stoke_space co-founders @AndyLapsa and @Rkt_Da to find out why they chose to take on one of the hardest problems in rocket science, how being a smaller startup gives them an advantage, and what full reusability could unlock for the future of spaceflight.

00:00 — Intro
01:16 — Stoke Space’s mission: Rapid reusability
02:18 — Why Second Stage capsules fail reentry
03:34 — Stoke Space’s stage 2 solution
05:30 — Reusability-First Design Philosophy
07:25 — Early Engine Development & Test Strategy
10:48 — Vertical Integration & Manufacturing
11:21 — Iteration Speed as a Competitive Advantage
12:29 — Software as Core Infrastructure
14:00 — Path to Orbit & Launch Operations
15:04 — How This Could Change The World

OpenAI for Healthcare:

ElevenLabs just solved speech transcription.

Their new model, Scribe v2, is SOTA in word error rate. It handles pauses, laughs, complex words with ease.

It's so good at segmenting speakers and made ZERO errors on the ultimate test: identical twin voices. 

How it works 👇