Log inSign up
Ravi Theja
1,411 posts
user avatar
Ravi Theja
@ravithejads
Applied AI @MistralAI Previously - @llama_index (LlamaIndex) Focused on LLMs, Agents, RAG, and fine-tuning LLMs.
SF
linkedin.com/in/ravidesetty/
Joined December 2010
849
Following
6,051
Followers

New to X?

Sign up now to get your own personalized timeline!

Create account

By signing up, you agree to the Terms of Service and Privacy Policy, including Cookie Use.

TermsยทPrivacyยทCookiesยทAccessibilityยทAds Infoยทยฉ 2026 X Corp.
Don't miss what's happening
People on X are the first to know.
Log inSign up
  • Pinned
    user avatar
    Ravi Theja
    @ravithejads
    Aug 5, 2024
    ๐Ÿ”ฅ Releasing BRAG: High-Performance RAG Model Trained In $25 ๐Ÿš€ Weโ€™re thrilled to announce the launch of our RAG models, collectively known as BRAGโ€”a series of SLMs (3 SLMs and 1 Ultra SLM) specifically trained for Retrieval Augmented Generation (RAG). 1๏ธโƒฃ โ BRAG-Qwen2-7b-v0.1
    66K
  • user avatar
    Ravi Theja
    @ravithejads
    Feb 4, 2024
    ๐Ÿš€ ๐‚๐จ๐ซ๐ซ๐ž๐œ๐ญ๐ข๐ฏ๐ž ๐‘๐ž๐ญ๐ซ๐ข๐ž๐ฏ๐š๐ฅ ๐€๐ฎ๐ ๐ฆ๐ž๐ง๐ญ๐ž๐ ๐†๐ž๐ง๐ž๐ซ๐š๐ญ๐ข๐จ๐ง (๐‚๐‘๐€๐†) โญ ๐Ÿค” 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
    60K
  • user avatar
    Ravi Theja
    @ravithejads
    Dec 5, 2023
    ๐ŸŒŸ๐Ÿ“˜ Excited to share the @OpenAI Cookbook for Evaluating RAG Systems with @llama_index ๐Ÿš€ ๐Ÿ” Explore the RAG system evaluation: 1๏ธโƒฃ Synthetic Data Generation with GPT-4 2๏ธโƒฃ Detailed Retrieval Evaluation (Featuring Hit Rate & MRR Metrics) 3๏ธโƒฃ In-depth Response Evaluation
    39K
  • user avatar
    Ravi Theja
    @ravithejads
    Feb 11, 2024
    ๐Ÿš€ ๐‡๐ข๐๐€: ๐€ ๐‡๐ข๐ž๐ซ๐š๐ซ๐œ๐ก๐ข๐œ๐š๐ฅ ๐‚๐จ๐ง๐ญ๐ž๐ฑ๐ญ๐ฎ๐š๐ฅ ๐€๐ฎ๐ ๐ฆ๐ž๐ง๐ญ๐š๐ญ๐ข๐จ๐ง ๐‘๐€๐† ๐Ÿ๐จ๐ซ ๐Œ๐š๐ฌ๐ฌ๐ข๐ฏ๐ž ๐ƒ๐จ๐œ๐ฎ๐ฆ๐ž๐ง๐ญ๐ฌ ๐๐€๐ŸŒŸ ๐Ÿ’ก The usual RAG approach struggles with retrieval accuracy when faced with massive indistinguishable documents comprising text, tables,
    30K
  • user avatar
    Ravi Theja
    @ravithejads
    Mar 1, 2024
    ๐Ÿ”ฅ Releasing Telugu Gemma finetuned models - Perhaps the most advanced open-source Telugu LLM models out there! ๐Ÿš€ 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
    37K
  • user avatar
    Ravi Theja
    @ravithejads
    May 15, 2024
    ๐Ÿ”ฅ Navarasa at Google I/O ๐Ÿ”ฅ ๐ŸŽ‰ I'm super thrilled to share that our work (with @ramsri_goutham) - Navarasa has been showcased at Google I/O! ๐ŸŒ Navarasa is a Gemma-finetuned Indic LLM covering 15 Indic languages (+ English). It was great to see indicchat developed by
    36K
  • user avatar
    Ravi Theja
    @ravithejads
    Feb 25, 2024
    ๐Ÿš€ ๐€๐œ๐ญ๐ข๐ฏ๐ž๐‘๐€๐†: ๐‘๐ž๐ฏ๐ž๐š๐ฅ๐ข๐ง๐  ๐ญ๐ก๐ž ๐“๐ซ๐ž๐š๐ฌ๐ฎ๐ซ๐ž๐ฌ ๐จ๐Ÿ ๐Š๐ง๐จ๐ฐ๐ฅ๐ž๐๐ ๐ž ๐ฏ๐ข๐š ๐€๐œ๐ญ๐ข๐ฏ๐ž ๐‹๐ž๐š๐ซ๐ง๐ข๐ง๐ ย ๐ŸŒŸ Retrieval-augmented generation (RAG) combines LLMs with external knowledge to create more accurate and relevant answers. However, RAG's
    32K
  • user avatar
    Ravi Theja
    @ravithejads
    Jun 23, 2024
    ๐Ÿ”ฅ 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. ๐Ÿ’ก The MoA approach layers different LLM agents, each building on the outputs of others to
    24K
  • user avatar
    Ravi Theja
    @ravithejads
    Oct 23, 2023
    ๐Ÿš€ 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.
    42K
  • user avatar
    Ravi Theja
    @ravithejads
    Mar 6, 2024
    ๐Ÿ”ฅ ๐‘๐ž๐ฅ๐ž๐š๐ฌ๐ข๐ง๐  ๐ˆ๐ง๐๐ข๐œ ๐†๐ž๐ฆ๐ฆ๐š 7๐/2๐ ๐ˆ๐ง๐ฌ๐ญ๐ซ๐ฎ๐œ๐ญ๐ข๐จ๐ง ๐ญ๐ฎ๐ง๐ž๐ ๐ฆ๐จ๐๐ž๐ฅ ๐จ๐ง 9 ๐ˆ๐ง๐๐ข๐š๐ง ๐‹๐š๐ง๐ ๐ฎ๐š๐ ๐ž๐ฌ โ€” ๐๐š๐ฏ๐š๐ซ๐š๐ฌ๐š ๐Ÿš€ We are thrilled to share ๐ŸŒŸ ๐๐š๐ฏ๐š๐ซ๐š๐ฌ๐š, a Gemma 7B & 2B instruction-tuned models in 9 Indian Languages - Perhaps
    31K
  • user avatar
    Ravi Theja
    @ravithejads
    Jun 10, 2024
    ๐ŸŒŸ Prometheus-2: An Open-Source Evaluator LM for Your RAG Application โœ… 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. ๐Ÿง‘โ€โš–๏ธ It acts as an LLM judge, good at
    28K
  • user avatar
    Ravi Theja
    @ravithejads
    Feb 13, 2025
    Career Update: I have recently wrapped up my journey at @llama_index, and it has been an amazing experience from start to finish! From contributing to the open-source ecosystem in the early days to joining full-time, it has been a privilege to help the developer community and
    15K
  • user avatar
    Ravi Theja
    @ravithejads
    Feb 7, 2024
    ๐Ÿšจ๐‘๐ž๐ฅ๐ž๐š๐ฌ๐ข๐ง๐  ๐ญ๐ก๐ž ๐“๐ž๐ฅ๐ฎ๐ ๐ฎ-๐‹๐ฅ๐š๐ฆ๐š-7๐-๐ฏ0-๐ˆ๐ง๐ฌ๐ญ๐ซ๐ฎ๐œ๐ญ ๐Œ๐จ๐๐ž๐ฅ ๐Ÿš€ 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. ๐Ÿ”ฅ Today,
    34K
  • user avatar
    Ravi Theja
    @ravithejads
    Jan 18, 2024
    โญ RAG VS FINE-TUNING: A CASE STUDY ON AGRICULTURE ๐ŸŒพ 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. ๐ŸŒ ๐Ÿ”
    29K