Ravi Theja @ravithejads Twitter profile | Pikagi

Pikagi

Ravi Theja

@ravithejads

4,118

Followers

692

Following

124

Media

1,088

Statuses

AI Engineer and Developer Advocate at @llama_index (LlamaIndex) Focused on RAG, agents, and fine-tuning LLMs.

Bangalore, India

https://t.co/5p2DXaFHxy

Joined December 2010

Don't wanna be here? Send us removal request.

Pinned Tweet

@ravithejads

Ravi Theja

4 months

🔥 𝐑𝐞𝐥𝐞𝐚𝐬𝐢𝐧𝐠 𝐍𝐚𝐯𝐚𝐫𝐚𝐬𝐚-2.0 - 𝐆𝐞𝐦𝐦𝐚 7𝐁/2𝐁 𝐈𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧 𝐭𝐮𝐧𝐞𝐝 𝐦𝐨𝐝𝐞𝐥 𝐨𝐧 15 𝐈𝐧𝐝𝐢𝐚𝐧 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞𝐬 🚀 We are thrilled to release 🌟 𝐍𝐚𝐯𝐚𝐫𝐚𝐬𝐚-2.0, a Gemma 7B & 2B instruction-tuned model in 15 𝐈𝐧𝐝𝐢𝐚𝐧

Tweet media one

17

36

148

Last Seen Profiles

@ano_ny_mouse

@NightmareBuzzBR

@gullskept

@RackssVLR

@danielcroe

@PenrhyncochFC

@0xBEW

@iassael

@cxllme01

@jennasvamp

@SegadorRoberto

@bbtmkn

@RecidiviaRPG

@RennesBusiness

@the_dinguskhan

@HardRockMagfr

@BlueSteelFl

@theryman

@HangTheLamps

@galery_basah10

@PconservateurQc

@engdaivityfess

@gunpolicy

@plutopools

@Negrita1987_

@MHeepow

@mekibertuah123

@JUmp

@bjpenndotcom

@IamaCamera2

@artixhyperr

@amerpipedream

@JadeEssence_

@theannatobin

@NoirsPhantasy

@nikitabagley

@ravithejads

Ravi Theja

6 months

🚀 𝐂𝐨𝐫𝐫𝐞𝐜𝐭𝐢𝐯𝐞 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 𝐀𝐮𝐠𝐦𝐞𝐧𝐭𝐞𝐝 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 (𝐂𝐑𝐀𝐆) ⭐ 🤔 RAG heavily relies on the relevance of retrieved-context, which can sometimes be fully correct or only partially correct, or incorrect. CRAG (Corrective Retrieval Augmented

Tweet media one

9

111

519

@ravithejads

Ravi Theja

8 months

🌟📘 Excited to share the @OpenAI Cookbook for Evaluating RAG Systems with @llama_index 🚀 🔍 Explore the RAG system evaluation: 1️⃣ Synthetic Data Generation with GPT-4 2️⃣ Detailed Retrieval Evaluation (Featuring Hit Rate & MRR Metrics) 3️⃣ In-depth Response Evaluation

Tweet media one

2

51

279

@ravithejads

Ravi Theja

6 months

🚀 𝐇𝐢𝐐𝐀: 𝐀 𝐇𝐢𝐞𝐫𝐚𝐫𝐜𝐡𝐢𝐜𝐚𝐥 𝐂𝐨𝐧𝐭𝐞𝐱𝐭𝐮𝐚𝐥 𝐀𝐮𝐠𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐑𝐀𝐆 𝐟𝐨𝐫 𝐌𝐚𝐬𝐬𝐢𝐯𝐞 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭𝐬 𝐐𝐀🌟 💡 The usual RAG approach struggles with retrieval accuracy when faced with massive indistinguishable documents comprising text, tables,

Tweet media one

0

65

277

@ravithejads

Ravi Theja

5 months

🔥 Releasing Telugu Gemma finetuned models - Perhaps the most advanced open-source Telugu LLM models out there! 🚀 Last week, @Google released the Gemma family of models (7B and 2B), featuring an exhaustive tokenizer(256K). A quick test of its tokenizing capabilities revealed

Tweet media one

18

35

274

@ravithejads

Ravi Theja

2 months

🔥 Navarasa at Google I/O 🔥 🎉 I'm super thrilled to share that our work (with @ramsri_goutham ) - Navarasa has been showcased at Google I/O! 🌐 Navarasa is a Gemma-finetuned Indic LLM covering 15 Indic languages (+ English). It was great to see indicchat developed by

Tweet media one

15

28

268

@ravithejads

Ravi Theja

5 months

🚀 𝐀𝐜𝐭𝐢𝐯𝐞𝐑𝐀𝐆: 𝐑𝐞𝐯𝐞𝐚𝐥𝐢𝐧𝐠 𝐭𝐡𝐞 𝐓𝐫𝐞𝐚𝐬𝐮𝐫𝐞𝐬 𝐨𝐟 𝐊𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞 𝐯𝐢𝐚 𝐀𝐜𝐭𝐢𝐯𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 🌟 Retrieval-augmented generation (RAG) combines LLMs with external knowledge to create more accurate and relevant answers. However, RAG's

Tweet media one

5

73

258

@ravithejads

Ravi Theja

9 months

🚀 Career Update Excited to announce that I'm joining the amazing @llama_index team! Super thrilled to work with @jerryjliu0 , @disiok , @seldo , @yi_ding , @LoganMarkewich & the team. Diving into RAG/Agents? Let's brainstorm & build. DMs open.

34

12

244

@ravithejads

Ravi Theja

1 month

🔥 Mixture-Of-Agents enhances LLM Capabilities This paper from @togethercompute introduces an approach Mixture-of-Agents (MoA) that harnesses the collective strengths of multiple LLMs. 💡 The MoA approach layers different LLM agents, each building on the outputs of others to

Tweet media one

2

56

241

@ravithejads

Ravi Theja

5 months

🔥 𝐑𝐞𝐥𝐞𝐚𝐬𝐢𝐧𝐠 𝐈𝐧𝐝𝐢𝐜 𝐆𝐞𝐦𝐦𝐚 7𝐁/2𝐁 𝐈𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧 𝐭𝐮𝐧𝐞𝐝 𝐦𝐨𝐝𝐞𝐥 𝐨𝐧 9 𝐈𝐧𝐝𝐢𝐚𝐧 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞𝐬 — 𝐍𝐚𝐯𝐚𝐫𝐚𝐬𝐚 🚀 We are thrilled to share 🌟 𝐍𝐚𝐯𝐚𝐫𝐚𝐬𝐚, a Gemma 7B & 2B instruction-tuned models in 9 Indian Languages - Perhaps

Tweet media one

17

49

227

@ravithejads

Ravi Theja

2 months

🌟 Prometheus-2: An Open-Source Evaluator LM for Your RAG Application ✅ Prometheus-2 shows a high degree of correlation and agreement with human evaluations, GPT-4, and Claude3, making it a dependable evaluator LM for RAG applications. 🧑‍⚖️ It acts as an LLM judge, good at

Tweet media one

2

59

222

@ravithejads

Ravi Theja

6 months

🚨𝐑𝐞𝐥𝐞𝐚𝐬𝐢𝐧𝐠 𝐭𝐡𝐞 𝐓𝐞𝐥𝐮𝐠𝐮-𝐋𝐥𝐚𝐦𝐚-7𝐁-𝐯0-𝐈𝐧𝐬𝐭𝐫𝐮𝐜𝐭 𝐌𝐨𝐝𝐞𝐥 🚀 Last week, we announced Telugu-LLM-Labs, a joint independent collaborative effort by me, and @ramsri_goutham , where we released datasets translated and romanized in Telugu. 🔥 Today,

Tweet media one

11

28

200

@ravithejads

Ravi Theja

6 months

⭐ RAG VS FINE-TUNING: A CASE STUDY ON AGRICULTURE 🌾 While numerous studies have explored RAG versus Fine-Tuning, @Microsoft 's latest paper with @OpenAI GPT4, Llama2-13B, and Vicuna stands out with its unique focus on agricultural data from the US, India, and Brazil. 🌍 🔍

Tweet media one

3

46

188

@ravithejads

Ravi Theja

13 days

🔥 Releasing GraphRAG implementation with @llama_index [in beta]. 🤔 RAG struggles with broad questions over a dataset, like “What are the main themes in the dataset?”, as it requires query-focused summarization (QFS) rather than simple retrieval. GraphRAG proposes an approach

@llama_index

LlamaIndex 🦙

13 days

As a special weekend release, we’ve implemented an initial working implementation of GraphRAG concepts in @llama_index [in beta]. GraphRAG from Microsoft has a lot of cool ideas, and we’ve implemented the following main concepts: 1. Graph Generation: Build graph and community

Tweet media one

4

65

320

7

49

185

@ravithejads

Ravi Theja

1 month

🔥 Releasing `llama-agents` and LlamaCloud (waitlist) 🚀 We’re thrilled to announce the release of `llama-agents`, our new open-source framework that simplifies the creation, iteration, and deployment of multi-agent AI systems, turning your agents into production-ready

Tweet media one

3

32

178

@ravithejads

Ravi Theja

1 month

🤔 Are your LLM applications relying on LLMs as evaluators? 🔥If so, the paper "Finding Blind Spots in Evaluator LLMs with Interpretable Checklists" by @sumanthd17 and team is essential reading. 💡 Their research discovered that LLMs failed to detect quality drops in over 50%

Tweet media one

3

33

155

@ravithejads

Ravi Theja

9 months

🚀Announcing @llama_index Workshop (Online) ⭐ Modules: 1. RAG Paradigm with LlamaIndex 2. RAG Use Cases 3. Open-source LLMs Customisation 4. Metadata in RAG 5. Evaluation 6. Fine-tuning Embeddings 7. Text2SQL Project: Fine-tuned RAG System Register:

Tweet media one

7

17

145

@ravithejads

Ravi Theja

7 months

🔥 🚀 Evaluating RAG with @OpenAI GPT-4 vs Open-Source Prometheus Model with @llama_index Curious about how the Open-source Prometheus model stacks up in evaluating Retrieval-Augmented Generation (RAG)? 🤔 We've put the open-source Prometheus model to the test against the

Tweet media one

5

25

125

@ravithejads

Ravi Theja

1 year

🎉 Excited to share that our team won $1000 from #GoogleCloud at the #Lightspeed #Warpspeed2023 #GenAI hackathon! 🚀 We've developed an AI Tutor for all-nighters & last-minute revisions, providing personalized tutoring. 📚🤖 (1/4)

13

10

123

@ravithejads

Ravi Theja

1 year

DocsQues - a tool that is built using @gpt_index , will help you answer queries on documents. Recruiters can use this tool to efficiently gather information about job applicants. Demos on querying multiple candidate resumes. (1/n) #AI #NLP #recruiters #ATS #Resume #ML #GPT3

11

10

121

@ravithejads

Ravi Theja

9 months

❓Wondered how to build RAG with open-source LLMs & embeddings in @llama_index ? 👉 Starting with @replicate hosted @MistralAI LLM & @huggingface embedding models. One can use various other LLMs on Replicate: 👇Dive into this 🧵 for other methods.

Tweet media one

2

22

119

@ravithejads

Ravi Theja

9 months

Struggling to decide between @OpenAI Assistant API and your RAG pipeline?🤔 🛠️Our latest toolkit enables clear benchmarking between OpenAI Assistant API and the RAG pipeline with @llama_index . 📄Check out docs:

Tweet media one

3

27

118

@ravithejads

Ravi Theja

9 months

🚀 Introducing Multi-Modal RAG with @llama_index : 1️⃣Input: Accepts both text and image data. 2️⃣Retrieval: Uses ada-002 & CLIP to retrieve text/image data. 3️⃣Synthesis: GPT-4V (LLaVa/Fuyu) crafts answer combining both modalities. Blogpost:

Tweet media one

3

27

115

@ravithejads

Ravi Theja

3 months

. @llama_index love from Eiffel Tower.

Tweet media one

9

5

101

@ravithejads

Ravi Theja

7 months

🔥 Propositions as Chunks to Improve RAG Performance with @llama_index ✨ 🔍 One crucial choice in RAG pipeline design is chunking: should it be passage level, sentence level, or token level? This choice significantly impacts your retrieval and response generation performance.

Tweet media one

5

16

95

@ravithejads

Ravi Theja

1 year

Enjoyed presenting ' @llama_index : Basics to Production' at @Accel Launchpad recently! Topics covered: 1. Abstractions & Indices 2. Use Cases 3. Open-Source Integration 4. Masking PII 5. Routing Queries 6. Text to SQL 7. Real-world applications ( @kar2905 , @dhruv___anand ) 8.

Tweet media one

1

9

90

@ravithejads

Ravi Theja

10 days

🔥 Happy to share that @ramsri_goutham and I presented at GoogleIO Connect yesterday on ‘Navarasa: Extending Google’s Gemma Model to 15 Indic Languages.’ We were amazed to see our work showcased in the keynote session as well. Fun fact: 2019: Google rejected my application to

Tweet media one

Tweet media two

4

9

87

@ravithejads

Ravi Theja

6 months

🚀 RAPTOR: RECURSIVE ABSTRACTIVE PROCESSING FOR TREE-ORGANIZED RETRIEVAL 🌟 Current RAG systems struggle with thematic questions that require information from various parts of a document/ text. For instance, answering "How did Cinderella reach her happy ending?" from the

Tweet media one

2

15

83

@ravithejads

Ravi Theja

10 months

Start of GenAI meet-up with @NirantK . Posting live demos here.

Tweet media one

13

5

79

@ravithejads

Ravi Theja

6 months

🌟 LLaMA Beyond English: An Empirical Study on Language Capability Transfer 🌍 With the increasing popularity of non-English specific Large Language Models (LLMs), this paper gives interesting insights for extending LLaMA models beyond the English language. 🔍 The usual process

Tweet media one

0

24

76

@ravithejads

Ravi Theja

2 months

🌟 Comprehensive RAG Benchmark (CRAG) 🔥 If you're looking to enhance your RAG skills, here's a dataset to start with and build upon. 📚 CRAG is a factual question-answering benchmark dataset with 4,409 question-answer pairs. 🌐 Includes mock APIs for simulating web and

Tweet media one

0

16

73

@ravithejads

Ravi Theja

11 months

Text2SQL + RAG in @llama_index Ever wondered about product reviews in SQL tables? Not just texts, but a goldmine of insights! The key? Text2SQL + RAG with @llama_index Dive into how we can decode these reviews in my latest post:

Tweet media one

0

9

73

@ravithejads

Ravi Theja

10 months

Presented 'Automatic Knowledge Transfer (KT) Video Generation of Code Bases using @llama_index ' at @pyconindia Delighted by the response! Many are eyeing implementation in their org. Missed it? Check our solutions. Blog: Code:

Tweet media one

2

11

72

@ravithejads

Ravi Theja

5 months

🤔 𝐈𝐬 𝐑𝐀𝐆 𝐠𝐨𝐢𝐧𝐠 𝐭𝐨 𝐝𝐢𝐞? 👉 With the release of Gemini 1.5 Pro and its 1M context, this question is buzzing in the community, especially for us at @llama_index , who are at the forefront of RAG frameworks. 👀 Many quick takes focus on the potential higher latency

4

17

67

@ravithejads

Ravi Theja

4 months

🔥 LlamaParse to build RAG over Complex PDFs with @llama_index 🤯 Building efficient RAG over complex PDFs containing text, tables, and images is challenging. LlamaParse helps here for you to parse documents and enhance retrieval and context with @llama_index . 🤔 Why choose

0

13

65

@ravithejads

Ravi Theja

1 year

Excited to announce the launch of DocsQues on @felvindotcom , a tool designed to make document querying effortless! Check it out now at 📄 🔍 Thank you @hargup13 for helping me to host it on @felvindotcom cc: @jerryjliu0

1

9

65

@ravithejads

Ravi Theja

1 year

I'm honored to be recognized as one of the Top 5 GenAI Scientists by the @AnalyticsVidhya at Data Hack Summit 2023. Immense thanks to my mentors @jerryjliu0 , @NirantK , and @phanisrikanth33 .

Tweet media one

Tweet media two

4

2

63

@ravithejads

Ravi Theja

1 month

At the @southpkcommons Bengaluru launch event yesterday. Had a fanboy moment of talking to @rsanghvi . Met lots of interesting folks whom we generally see here - @1littlecoder , @anshubhide123 , @AashaySachdeva and many more.

Tweet media one

Tweet media two

Tweet media three

2

1

64

@ravithejads

Ravi Theja

4 months

🚀 Deploying Navarasa-2.0 - a Gemma finetuned model in 15 Indian Languages 🌟 On March 18th, 2024, we proudly unveiled Navarasa-2.0 - a cutting-edge Gemma 7B/2B instruction-tuned model proficient in 15 Indian languages (+ English). We are thrilled to observe the growing

Tweet media one

7

19

62

@ravithejads

Ravi Theja

25 days

🤔 How did I end up at @llama_index ? This is a question I often receive at hackathons, workshops, and talks. Thanks to @1littlecoder for featuring me on his podcast, where I discussed: 1⃣Open-source software contributions. 2⃣Taking risks. 3⃣Undertaking new projects and

@1littlecoder

1LittleCoder💻

25 days

🔥 I was lucky speak to the super dynamic @ravithejads about: 🚀 OSS Contributions 🚗 Path to become an AI Engineer 🔥 Characteristics to succeed Of course, some AI stuff! Full interview👇🏾

4

4

39

4

6

62

@ravithejads

Ravi Theja

6 months

💡ChatQA: Building GPT-4 Level Conversational QA Models Builds ChatQA conversational QA models, that obtain OpenAI GPT-4 level accuracies. 🌟 Key Contributions: 1️⃣ Two-Stage Instruction Tuning Method: - Stage 1: Supervised Fine-Tuning on a mix of instruction-following and

Tweet media one

4

15

63

@ravithejads

Ravi Theja

1 year

Confused about using @llama_index indices or configuring parameters for building robust QA Systems? See my new video on: 1️⃣ VectorStore Index 2️⃣ Storage Context 3️⃣ Service Context 4️⃣ List Index 5️⃣ KeyWordTable Index 📺 Video: 📔 Notebook:

Tweet media one

2

11

61

@ravithejads

Ravi Theja

7 months

🌟Reflecting on a Transformative 2023 The year 2023 marked a significant transformation in my professional life, beginning with a simple interaction with the @LlamaIndex handle on the @X platform. 1⃣Starting open source with @LlamaIndex 2⃣Assisted multiple companies in building

5

4

60

@ravithejads

Ravi Theja

9 months

🚀 Launching create-llama Now launch chatbots on your docs with create-llama - a simple CLI to generate a full-stack @llama_index application using Python or TS. All you need to do from CLI is run: npx create-llama Video: Blog:

Tweet media one

2

12

59

@ravithejads

Ravi Theja

1 year

Enjoyed presenting 'Beyond the Basics: Leveraging @llama_index from Concept to Production' at the @SpecialeInvest GenAI meetup in Chennai recently. 1. Abstractions and Indexing 2. Use-Cases 3. Open-Source Integration 4. PII Masking 5. Query Engine 6. Chat Engine 7. Routing 8.

2

9

58

@ravithejads

Ravi Theja

10 months

RAG (Retrieval Augmented Generation) apps need strong retrieval, reliant on quality embeddings. @OpenAI embeddings API - text-embedding-ada-002 holds the 13th spot on MTEB. The last major update was in Dec'22. Is a new @OpenAI embedding API round the corner? 🤔

Tweet media one

3

2

57

@ravithejads

Ravi Theja

1 year

Unleash the potential of @llama_index with my latest tutorial, complete with dynamic visuals! You'll gain insights into: 1. Key components of LlamaIndex 2. Building QA System with High-Level Components 3. Building QA System with Low-Level Components Link:

@jerryjliu0

Jerry Liu

1 year

What are *all* the key components needed to build an LLM question-answering app? A picture is worth a thousand words 🖼️, and @ravithejads provides amazing graphics in our brand-new tutorial covering everything you need to know about RAG + @llama_index :

1

29

144

2

7

55

@ravithejads

Ravi Theja

10 months

🚀 Decoding the perfect chunk size for RAG Systems using @llama_index ⚙️ 🤔 How does chunk size impact RAG system efficiency & accuracy? Discover the balance 🎯 in our latest blog post using the Response Evaluation module of @llama_index :

Tweet media one

2

5

56

@ravithejads

Ravi Theja

10 months

Discover @llama_index Templates on @replit 🦙 Dive straight into building LLM apps using our templates on Replit. No setup hassles, just instant app creation! Explore ready-to-use @streamlit apps for QA and Summarization. QA Template: Summarization

0

5

53

@ravithejads

Ravi Theja

30 days

🔥 Gemma-2 is Here with 9B and 27B Models! What sets Gemma-2 apart? 1️⃣ Peak Performance: The 27B model outshines competitors, even those twice its size, while the 9B version leads its class. 2️⃣ Cost Efficiency: Operates on single devices like Google Cloud TPU and NVIDIA GPUs,

Tweet media one

5

8

52

@ravithejads

Ravi Theja

27 days

🔥 Hackathon Alert! We’re thrilled to partner with @Stellaris_VP for the AI Agents Hackathon this weekend, July 6-7. 🏆 Special prize for the best project featuring llama-agents and @llama_index . 📝 Register here: 🔍 Haven’t explored our llama-agents

Tweet media one

3

11

53

@ravithejads

Ravi Theja

1 year

🚀Excited to share that I recently had the opportunity to speak at the @hasgeek Generative AI meetup on "LlamaIndex ( @gpt_index ) and Evaluating QA Systems using LlamaIndex"! (1/3) Slides deck: Talk: cc: @jerryjliu0

Tweet card media

Evaluation of QA systems built on LlamaIndex

This talk was delivered at the April 2023 Generative AI meetup by Ravi Theja, data scientist at Glance (InMobi). Details of the talk and link to slides for t...

www.youtube.com

4

9

48

@ravithejads

Ravi Theja

11 months

📢 Exciting Workshop Alert on @llama_index ! Thrilled to announce that I'll be conducting a workshop on @llama_index with @hasgeek this Saturday, August 26th, from 10 AM - 3 PM. 🚀 Topics Covered: 1️⃣ Understanding the Retrieval Augmented Generation (RAG) paradigm 2️⃣ Indexing

Tweet media one

3

9

46

@ravithejads

Ravi Theja

1 year

Had an amazing time presenting @llama_index and its applications - Albus ( @kar2905 ) & LlamaReaders ( @dhruv___anand ) at @scaletogether last week. Thanks, @177pc , for the opportunity. Looking forward to future sessions. PC: @s16r442

Tweet media one

3

5

45

@ravithejads

Ravi Theja

3 months

Had an amazing time traveling around Amsterdam, The Hague, Brussels, and Paris. Now, it's back to work—excited to ship something soon! Also looking forward to some interesting updates from @OpenAI and @Google in the next couple of days.

Tweet media one

Tweet media two

Tweet media three

Tweet media four

5

0

44

@ravithejads

Ravi Theja

1 year

VideoQues - a tool that is built using @gpt_index and @OpenAI whisper, will help you answer queries on videos. Demos here are shown specifically on some of the YouTube videos. Demo on a YT Video - @TEDTalks of @elonmusk (1/n)

3

3

41

@ravithejads

Ravi Theja

11 months

🚀 Production-Ready RAG App with @llama_index 🦙🚀 Transforming a mere prototype into a full-blown production solution is no small feat and @llama_index has shown the way, doing it with unmatched flair! Checkout - a full-stack, production-ready RAG app!

@llama_index

LlamaIndex 🦙

11 months

We’re excited to open-source - a full-stack, production-ready RAG app! 🦙🏦 Supports streaming, reasoning steps, citations, intuitive UI This can save you weeks/months of hard work in trying to build a prod LLM app from scratch🔥

10

101

568

1

5

41

@ravithejads

Ravi Theja

1 month

🔥Fine-Tuning @MistralAI Models and RAG Using @llama_index 🚀 @MistralAI has recently launched a fine-tuning API that enables users to efficiently enhance their models. 👉Our new cookbook illustrates how to fine-tune the open-mistral-7b model using the MistralAIFinetuneEngine to

Tweet media one

1

3

39

@ravithejads

Ravi Theja

9 months

Want to use your own data stored in VectorDB with @OpenAI Assistant API? We've added OpenAIAssistantAgent to enable Assistant API with your VectorDB! Follow this thread for details on using @llama_index OpenAIAssistantAgent with Assistant API on your own indexed knowledge.

@jerryjliu0

Jerry Liu

9 months

Want to use @OpenAI Assistants API but already using a vector DB? 🤔 No problem. With our brand-new `OpenAIAssistantAgent` abstractions ✨ we show you how to “Bring Your Own Vector Store” (BYOVS) 🍾 You don’t have to be locked in to the retrieval API! Here’s a full guide on

Tweet media one

Tweet media two

9

65

454

3

5

39

@ravithejads

Ravi Theja

9 months

🚀Parallel Function Calling With @llama_index @llama_index enhances its capabilities with @OpenAI new Parallel Function Calling. Now, multiple function calls and diverse structured data extraction are possible from just one prompt. Notebook:

Tweet media one

2

8

37

@ravithejads

Ravi Theja

1 year

Huge Congratulations @jerryjliu0 , @disiok , and @llama_index team on the seed funding by @GreylockVC ! As an early contributor, thrilled to see the growth. Excited to add many more interesting features, enriching the framework's capabilities and impact on private data management.

@jerryjliu0

Jerry Liu

1 year

I’m super excited to make it official: @disiok and I have started a company around @llama_index , and we’ve raised a $8.5M seed round led by @GreylockVC ! 🔥🚀 We are building the open-source data framework to unlock LLM capabilities on your private data.

96

119

1K

0

1

37

@ravithejads

Ravi Theja

8 months

🚀Launching @llama_index 0.9v Check out the top 3 features: 1⃣Incremental Data Updates 2⃣IngestionPipeline for production-ready RAG 3⃣Transformation Caching for efficiency Let's dive into each of these in this 🧵👇

Tweet media one

1

7

36

@ravithejads

Ravi Theja

1 year

LlamaIndex @gpt_index now has a new QueryEngine - SQLAutoVectorQueryEngine, a potent combination of SQL and vector databases for enriched insights over both structured and unstructured data. A thread on the new feature 🧵💡🚀 (1/7) #LlamaIndex #LLM #AI #SQL #Vectordb

@jerryjliu0

Jerry Liu

1 year

There are two separate LLM stacks over structured/unstructured data: - Text-to-SQL (structured) - Vector db QA (unstructured) Introducing a new *combined* stack: use SQL for structured analytics, augment w/ unstructured context from a vector db! 🔥

12

93

517

1

2

36

@ravithejads

Ravi Theja

1 month

Had a wonderful time talking about @llama_index at @iimb_official InnovAI - @e2enetworks Conference. It was a fantastic opportunity to engage with hackers and founders throughout the event. I also had the pleasure of attending @velumania talk in the evening, which was truly

Tweet media one

0

4

34

@ravithejads

Ravi Theja

1 year

Late night hacks at @lightspeedvp #Warpspeed2023 #GenAI hackathon. @ojasvi_yadav @AmirNagri @sumodm @AyushDeva #hackathon #genai #ai #lightspeed

Tweet media one

Tweet media two

0

3

35

@ravithejads

Ravi Theja

6 months

🚀 𝐄𝐧𝐡𝐚𝐧𝐜𝐢𝐧𝐠 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 𝐏𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞 𝐰𝐢𝐭𝐡 𝐀𝐥𝐩𝐡𝐚 𝐓𝐮𝐧𝐢𝐧𝐠 𝐢𝐧 𝐇𝐲𝐛𝐫𝐢𝐝 𝐒𝐞𝐚𝐫𝐜𝐡 𝐢𝐧 𝐑𝐀𝐆 𝐰𝐢𝐭𝐡 @llama_index 🌟 🤔 Ever wondered if your RAG application is effective for all kinds of user queries? For different user queries, a

Tweet media one

1

13

35

@ravithejads

Ravi Theja

5 months

🚀 𝐓𝐎𝐎𝐋𝐕𝐄𝐑𝐈𝐅𝐈𝐄𝐑: 𝐆𝐞𝐧𝐞𝐫𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐭𝐨 𝐍𝐞𝐰 𝐓𝐨𝐨𝐥𝐬 𝐯𝐢𝐚 𝐒𝐞𝐥𝐟-𝐕𝐞𝐫𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 🌟 Despite progress in teaching LLMs to use specific tools, mastering new, unseen tools poses a significant challenge. Traditional methods, relying on

Tweet media one

0

9

33

@ravithejads

Ravi Theja

9 months

🎉 @llama_index turns 1 🎂 From @jerryjliu0 weekend project to massive growth: 1️⃣450+ contributors 2️⃣3,000+ projects using our library 3️⃣4,000 Discord members 4️⃣47K lines of Python code 5️⃣900K monthly downloads Year's highlights in our blog:

Tweet media one

1

1

34

@ravithejads

Ravi Theja

4 months

🌟 RAG in 2024 with @llama_index Super excited to share that I will be speaking about RAG in 2024 with @llama_index at @SaamaOfficial Connect March Meetup in Chennai this Saturday. I will be covering a wide range of topics: 1️⃣ Retrievers 2️⃣ Agents 3️⃣ Latest Research in RAG 4️⃣

Tweet media one

0

5

33

@ravithejads

Ravi Theja

2 years

I am super proud of the contribution that I have made to @gpt_index over the past few weeks. Now you can query "MULTIMODAL" data - pdfs 📝, docx 📄, images(png, jpeg, jpg) 🖼️, audio(mp3) 🎵 and video(mp4) 🎬 files using @gpt_index . Go check out this thread. cc: @jerryjliu0

@jerryjliu0

Jerry Liu

2 years

A direction of @gpt_index that I’ve become increasingly excited about is its *multimodal* ingestion. Starting to dramatically simplify time to value from unstructured data. We can now parse .pdf 📝, .docx, .jpg, .png 🖼️, and even .mp3 💬 and .mp4 🎬 files into Document objs!🧵

12

30

294

2

0

31

@ravithejads

Ravi Theja

2 months

@mariaKhalusova We ( @llama_index ) have created a tutorial series to get started with LLM Agents and RAG. You can check them out here -

@jerryjliu0

Jerry Liu

3 months

Check out this 7-part tutorial series by @ravithejads on an introduction to agents. Agents can be really complicated, but if you're trying to get a basic handle of what agents are and how to use them, this is a great intro. ✅ Learn how basic agent implementations work

3

68

295

2

4

32

@ravithejads

Ravi Theja

9 months

🚀 Excited to share that I will be speaking about - our work on analyzing SEC 10-K documents using @llama_index at the @LightningAI meetup tomorrow, November 7th! Register:

Tweet media one

0

2

33

@ravithejads

Ravi Theja

5 months

🔥 Network of RAGs with @llama_index 🌐 🤔 Are you in an enterprise setting where RAG pipelines are developed by various organizations, and you're looking to leverage them effectively? 🚀 𝐥𝐥𝐚𝐦𝐚-𝐢𝐧𝐝𝐞𝐱-𝐧𝐞𝐭𝐰𝐨𝐫𝐤𝐬 – will simplify this integration process! Let's

Tweet media one

0

6

33

@ravithejads

Ravi Theja

10 months

5th Edition of @llama_index Update Series! 🌐 is now OPEN-SOURCE! 🔥 @Replit templates - Kickstart your dev projects. 📚 Retrieval Tips for RAG, Fine-tuning & Build-from-Scratch Guides. ⬆️ Packed with feature enhancements, 🤝 collaborations, 📝

3

3

30

@ravithejads

Ravi Theja

10 months

📄Diving into the @MistralAI 7B paper, I came across something interesting! When asked, "How to kill a Linux process?" with system prompts activated, Llama 2 chose not to respond, while @MistralAI 7B gave an accurate answer.🔥 Some highlights from the paper👇

Tweet media one

4

6

30

@ravithejads

Ravi Theja

2 months

We'll be talking about @llama_index and Navarasa with @ramsri_goutham and Manish tomorrow. Join us if you're interested in LLMs, RAG, or Agents! Checkout for more details👇

@lucifer_x007

Manish Sharma 📊 - e/acc 🤖

2 months

Twitter Space Announcement - 1 Excited to share that we will be having @ravithejads and @ramsri_goutham on our space-1. Ravi Theja - DevRel @ @llama_index . Co-Maintainer of Navrasa2.0 Ramsri Goutham - Maintainer of 2 Big SaaS Startup. Co-Maintainer of Navrasa2.0. Date - June

6

2

32

0

8

29

@ravithejads

Ravi Theja

1 year

I recently had the pleasure of delivering a talk on "Building QA Bots by Connecting LLMs with External Data Sources Using @gpt_index / @LangChainAI " at the Machine Perception Lab at my alma mater - @IIITB_official .(1/3) #AI #ML #LLM #NLP #GPTIndex #LangChain

Tweet media one

Tweet media two

4

2

27

@ravithejads

Ravi Theja

11 months

📣 PyCon India - @llama_index Workshop 🦙 Thrilled to share that I will be hosting a hands-on workshop at @pyconindia on Retrieval Augmented Generation (RAG) with @llama_index 🦙. 🎯 What to Expect? 1️⃣ Dive into LLMs and the RAG paradigm. 2️⃣ Explore LlamaIndex components. 3️⃣

Tweet media one

1

6

28

@ravithejads

Ravi Theja

1 year

Engaging panel discussions and amazing demos across the RAG, Infrastructure, and Fashion domains. Loved seeing my 'India's AI Dream' art on display. Each meetup stokes my passion for making new connections. Big thanks to @aishwarya_08 , @Inferless_ , @huggingface ,

Tweet media one

Tweet media two

Tweet media three

Tweet media four

0

1

26

@ravithejads

Ravi Theja

6 months

🔥 🚀 Telugu-LLM-Labs will be releasing the TinyLlama-1.1B-Telugu-Romanization-Base and TinyLlama-1.1B-Telugu-Romanization-Instruct models next week 👇

@ramsri_goutham

Ramsri Goutham Golla

@ramsri_goutham

6 months

Open-source models like Llama2 and Mistral are not good with Indian languages because of the scarcity of native tokens in the tokenizer and the lack of a strong presence in training data. One option is to expand the tokenizer vocabulary and pre-train with a lot of native text.

Tweet media one

7

3

54

0

1

27

@ravithejads

Ravi Theja

29 days

🔥llama-agents from @llamaindex are here! What are you hacking on with llama-agents this weekend? As a @llama_index user, you should try out LLAMA-CARD-GAME. 🎮 I’m teaming up with @phanisrikanth33 for a fun-packed hack session! ⌨️ Haven’t checked out llama-agents yet?

Tweet media one

1

5

26

@ravithejads

Ravi Theja

21 days

🔥 9600x speedup in search latency! 🚀From 8 minutes to 50ms. The tricks that @NirantK brings to the table are amazing. Here’s why you should consider attending his seminars on `Search for RAG Apps` in July for Sr. Engineers & CTOs building RAG for the first time! Check it

@NirantK

Nirant

24 days

Played tech support to my friend @akshay_pachaar — his @qdrant_engine latency went from 8 minutes down to 50ms. That's a 9600x speedup — that's how search engineering is done!

Tweet media one

6

2

63

0

6

26

@ravithejads

Ravi Theja

1 year

📣 Introducing @llama_index Update blog series! Stay updated, engaged, & inspired with the evolving open-source project. It covers everything—from new features to webinars, tutorials, hackathons, and community events. Check out the first update:

Tweet card media

LlamaIndex Update — 6/26/2023

Greetings, LlamaIndex community!

@llama_index

LlamaIndex 🦙

1 year

Introducing: the LlamaIndex Update Series 🦙✨ We recap feature releases, tutorials, webinars, and events in one central place, at a regular cadence. S/o @ravithejads In the last two weeks: 💡 19 features 🧑‍🏫 >= 5 tutorials 🎊lots of hackathons/events

0

6

40

0

2

26

@ravithejads

Ravi Theja

1 year

Second blog in @llama_index Update series is out! 🔧 Explore LlamaIndex 0.7.0 👀 See “RecursiveRetriever” for hierarchic data 🤝With @anyscalecompute , @chainlit_io , @weaviate_io , @AirbyteHQ 🔍 Structured data extraction & updated data model Read:

Tweet card media

LlamaIndex Update — 07/10/2023

Greetings once again, LlamaIndex community!

@llama_index

LlamaIndex 🦙

1 year

If you want to catch up on our updates in the last two weeks, check out our recap (s/o @ravithejads ): 🔧 LlamaIndex 0.7.0 🙌 Content with @anyscalecompute , @chainlit_io , @weaviate_io , @AirbyteHQ , etc. 🧩 Structured data extraction 💾 Updated data model

0

7

33

1

6

26

@ravithejads

Ravi Theja

1 year

Excited to share my first article on the #LlamaIndex blog! 🎉 Diving into the proper evaluation of LLM-powered QA systems and discussing the features that came from our valuable enterprise user feedback. Check it out here:

Tweet card media

Building and Evaluating a QA System with LlamaIndex

Introduction

@jerryjliu0

Jerry Liu

1 year

LlamaIndex now has a blog! 🦙📝 The first article is on proper evaluation of LLM-powered QA systems, by @ravithejads . Ravi helped to contribute most of the evaluation modules within LlamaIndex - a lot of the features we offer is the result of

5

28

163

3

2

24

@ravithejads

Ravi Theja

8 months

🔥Multi-Modal RAG with GPT4-V and Table Transformer in @llama_index 📄Handling tables in PDF documents is a huge challenge for RAG systems. 🔎However, Microsoft's new Table Transformer detects tables in document images. 🤝We explored pairing Table Transformer with @OpenAI

Tweet media one

0

3

25

@ravithejads

Ravi Theja

7 months

🔥 LlamaDatasets - Explore and Benchmark RAG Pipelines with @llama_index 🤔 Are you looking for datasets to enhance your exploration and benchmarking of RAG pipelines? 🚀 Head over to LlamaDatasets – for building, experimenting, and benchmarking RAG pipelines. It's a simple

Tweet media one

2

7

23

@ravithejads

Ravi Theja

1 year

Seems like text-davinci-002-20221122 is @OpenAI ChatGPT API.

Tweet media one

5

0

24

@ravithejads

Ravi Theja

1 year

Excited to present "Building QA Systems with @llama_index at @NexusVP X @TheProductfolks GenAI Rush, this Sun, July 16th at 8 PM. Covering Indices, QueryEngines, RouterEngines, Text2SQL & more. Join the final GenAI hackathon on July 29. Register:

Tweet media one

1

2

23

@ravithejads

Ravi Theja

10 months

📢 . @llama_index Weekly Newsletter Here are the key highlights. 1️⃣ @arizeai Phoenix integration. 2️⃣ @huggingface Embeddings support. 3️⃣ Rolled out Multi-Document Agents. 4️⃣ @streamlit Hackathon winners - NewsGPT & FinSight. Checkout:

Tweet media one

1

1

23

@ravithejads

Ravi Theja

1 year

Exciting news🚀 I've built HolyBot that can provide answers and relevant verses for your queries based on Bhagwad Gita, Quran, or Bible, built using @OpenAI ChatGPT and @pinecone . HuggingFace Space: Github Code base: #LLM #AI

1

0

21

@ravithejads

Ravi Theja

1 year

Tutorial coming up on Question Answering system combining @gpt_index , @qdrant_engine , and Cohere rerank ( @CohereAI ). Stay tuned for the blog post and Colab notebook.

@NirantK

Nirant

1 year

Work in Progress: Sketch for a Question Answering System Tutorial What is less than clear to you? cc @ravithejads

Tweet media one

3

1

31

1

1

21

@ravithejads

Ravi Theja

1 year

LlamaIndex ( @llama_index ) is sponsoring @Stellaris_VP and @join_ef GenAI Hackathon this weekend. Cash prizes to the top 3 best teams that develop apps using LlamaIndex. I'll be there in person to assist. more details:

Tweet card media

GenAI Hackathon: AI for India | 10-11 Jun | Bengaluru | Registrations closed · Luma

Join us for the GenAI Hackathon around the theme of 'AI for India'. This hackathon challenge is all about using AI to solve real problems that are unique to…

@llama_index

LlamaIndex 🦙

1 year

Excited to be sponsoring the GenAI Hackathon hosted by @Stellaris_VP and @join_ef in Bengaluru, India! 🦙 If you're in the area, register here: @ravithejads will be around to help!

Tweet media one

0

1

21

0

4

21

@ravithejads

Ravi Theja

1 year

📣 The 3rd Edition of the @llama_index Update Series is live! Dive into: 🚀 Enhanced Data Agents 💡 Upgraded LlamaIndex TS 🔍 New Benchmarking 🌟 Tutorials, Blogs, Demos, and Events Get all the details:

0

3

19

@ravithejads

Ravi Theja

1 year

@danshipper Use Azure OpenAI, they offer GPT-4-32K.

4

0

19

@ravithejads

Ravi Theja

1 year

Added the QueryResponseEvaluator module to the existing Evaluation framework on LlamaIndex @gpt_index . This evaluates the alignment of the response generated from LLM with the Query and context source information. #LlamaIndex #Evaluation #LLM #AI

@jerryjliu0

Jerry Liu

1 year

To evaluate retrieval-augmented LLM’s, we can compare the query/response/source context to see if they match. We just added this to @gpt_index 🔥. Best part, it’s label-free! A “correct” result is when both the response and source matches the query. Check out details below 🧵

Tweet media one

4

18

152

1

1

18

@ravithejads

Ravi Theja

1 year

🔍 Wondering about @llama_index ? Tomorrow, I'm speaking at the @hasgeek @fifthel conference on "Why @llama_index : Unlocking the Potential of LLM's". Dive deep with us & unlock the power of LLM's! Register & explore: 📢

Tweet media one

1

5

19

@ravithejads

Ravi Theja

1 year

LlamaIndex ( @llama_index ) is at MumbaiHacks ( @mumbai_tech_ ) this weekend. $500 bounty for the best apps developed using LlamaIndex. I'll be there in person to assist and can't wait to see what you build. Register here:

@llama_index

LlamaIndex 🦙

1 year

The 🦙 is going international 🌏. First stop: Mumbai, India! If you're around, come check out MumbaiHacks this weekend ( @mumbai_tech_ ). We're sponsoring a $500 USD bounty for apps w/ LlamaIndex 🎁. @ravithejads will be there to help 💪 Register here:

1

1

21

0

2

19

@ravithejads

Ravi Theja

8 months

🚀 Evaluate Multi-Modal RAG Pipeline with @llama_index Evaluation at Retrieval and Generation stages. 1️⃣ Image Retrieval - Hit-rate & MRR metrics. 2️⃣ Generation from Images + Text - Assessing faithfulness, relevance, and accuracy metrics. Blog Post:

Tweet media one

1

4

19

@ravithejads

Ravi Theja

4 months

💡JavaScript RAG Web Apps with @llama_index 📝 Learn how to develop a RAG web application using LlamaIndex with the latest course from @DeepLearningAI . 1️⃣ Develop a RAG web application for efficient data querying. 2️⃣ Craft an agent that integrates with multiple data sources,

0

4

17