<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by KARJAL PRIYANKA on Medium]]></title>
        <description><![CDATA[Stories by KARJAL PRIYANKA on Medium]]></description>
        <link>https://medium.com/@karjalpp?source=rss-3d3ccfe5cc64------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*SJZ4nwmjHZ1BKjWT</url>
            <title>Stories by KARJAL PRIYANKA on Medium</title>
            <link>https://medium.com/@karjalpp?source=rss-3d3ccfe5cc64------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 08:11:24 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@karjalpp/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Bharat Gen vs Global AI Giants: How India Is Building Its Own GPT!]]></title>
            <link>https://medium.com/@karjalpp/bharat-gen-vs-global-ai-giants-how-india-is-building-its-own-gpt-026b48c158a1?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/026b48c158a1</guid>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Wed, 18 Jun 2025 12:47:21 GMT</pubDate>
            <atom:updated>2025-06-18T15:54:20.875Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>BharatGen vs Global AI Giants: How India Is Building Its Own GPT!</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/715/1*4frrSEyOvxQdbzkZC4sncA.png" /></figure><h3><strong>Introduction:</strong></h3><p>India is a land of over <strong>1.4 billion people</strong> who speak <strong>22 official languages</strong> and hundreds of dialects. Despite the massive growth of AI globally, many models fall short when it comes to understanding and generating content in <strong>Indian languages</strong> like Hindi, Tamil, Telugu, Bengali, and others.</p><p>That’s where <strong>Bharat Gen</strong> comes in — a generative AI initiative built by India, for India. This article explores what Bharat Gen is, how it’s being used to empower multilingual AI applications, and how developers can get started with it today using open-source tools.</p><p>🌐 What Is Bharat Gen?</p><p><strong>Bharat Gen</strong> is India’s effort to develop <strong>multilingual large language models (LLMs)</strong> tailored for the country’s diverse linguistic and cultural landscape. It is part of broader initiatives like <strong>Bhashini</strong> and <strong>IndiaAI</strong>, supported by the Government of India and organizations like <strong>AI4Bharat</strong>, <strong>IIT Madras</strong>, and <strong>Sarvam AI</strong>.</p><h3>✅<strong> Key Highlights of Bharat Gen:</strong></h3><ul><li>Supports <strong>multiple Indian languages</strong></li><li>Open-source models available via Hugging Face</li><li>Enables <strong>chatbots</strong>, <strong>translation</strong>, <strong>Q&amp;A</strong>, <strong>summarization</strong>, and more</li><li>Used in <strong>rural helpdesks</strong>, <strong>e-governance</strong>, and <strong>education</strong></li></ul><h3><strong>🔧 Getting Started with Bharat Gen:</strong></h3><p>Most of Bharat Gen’s models are available on <strong>Hugging Face</strong>, especially under the AI4Bharat initiative. Let’s walk through a simple example using their <strong>Hindi language model</strong>.</p><p><strong>🧪 Code Example: Generating Hindi Text Using Bharat Gen:</strong></p><p>Let’s generate some Hindi text using the IndicGPT-Hindi model.</p><pre># Install the required libraries<br>!pip install transformers torch --quiet<br><br># Load model and tokenizer<br>from transformers import AutoModelForCausalLM, AutoTokenizer<br>import torch<br><br>model_name = &quot;ai4bharat/IndicGPT-Hindi&quot;<br>model = AutoModelForCausalLM.from_pretrained(model_name)<br>tokenizer = AutoTokenizer.from_pretrained(model_name)<br><br># Input text in Hindi<br>input_text = &quot;भारत एक महान देश है। यहाँ के लोग&quot;<br><br># Tokenize and generate<br>inputs = tokenizer(input_text, return_tensors=&quot;pt&quot;)<br>outputs = model.generate(**inputs, max_new_tokens=50, do_sample=True)<br><br># Decode and print<br>generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)<br>print(generated_text)</pre><p>✅ <strong>Expected Output (Example)</strong>:</p><pre>भारत एक महान देश है। यहाँ के लोग अनेक भाषाएँ बोलते हैं, विविधता में एकता का परिचय देते हैं, और संस्कृति<br> की धरोहर को सहेजते हैं।</pre><h3>📦 Where to Find Bharat Gen Models</h3><p>You can find Bharat Gen models on:</p><ul><li>🔗 Hugging Face — AI4Bharat Collection</li><li>🔗 <a href="https://bhashini.gov.in/">Bhashini Developer Portal</a></li><li>🔗 <a href="https://indiaai.gov.in/">IndiaAI</a></li></ul><p>📢 Why Bharat Gen Matters</p><p>Unlike Western models that are primarily English-centric, Bharat Gen is <strong>inclusive</strong>, <strong>culturally aligned</strong>, and <strong>linguistically grounded</strong>. It ensures that <strong>rural users, students, and local enterprises</strong> can benefit from AI in a language they understand and trust.</p><p>🚀 What’s Next?</p><p>The future of Bharat Gen involves:</p><ul><li><strong>Multimodal models</strong> (image, voice, text)</li><li>Expansion to <strong>more dialects</strong></li><li>Integration with <strong>India Stack</strong>, e.g., Aadhaar, UPI, and DigiLocker</li></ul><p>🧩 Final Thoughts</p><p>Bharat Gen is more than just a model — it’s a mission to democratize AI for <strong>every Indian</strong>, regardless of the language they speak. By using these models, developers, startups, and enterprises can build solutions that are <strong>truly inclusive</strong> and <strong>deeply impactful</strong>.</p><p>If you’re working on voicebots, citizen portals, regional learning apps, or AI for Bharat — this is your call to action.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/873/1*GKLhx-Ay4dYlyzbnwSxSyg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ln3wXn2d1i3x-ptnLVTVdQ.png" /></figure><h3>💬 Conclusion:</h3><p>While GPT, Claude, and DeepSeek are <strong>global LLMs</strong>, <strong>Bharat Gen</strong> fills a <strong>critical local gap</strong> by focusing on <strong>regional languages and cultural relevance</strong>. It’s not just about matching global AI standards — it’s about building <strong>AI for Bharat</strong>.</p><h3>🔄 <strong><em>Let’s not just build in India. Let’s build for Bharat.</em></strong></h3><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Dd3-UyoBbtwwtKpQ" /><figcaption>Photo by <a href="https://unsplash.com/@wizwow?utm_source=medium&amp;utm_medium=referral">Donald Giannatti</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=026b48c158a1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“Linear vs. Loopy: Choosing Between LangChain and LangGraph in 2025”]]></title>
            <link>https://medium.com/@karjalpp/linear-vs-loopy-choosing-between-langchain-and-langgraph-in-2025-2e27cff7bd9c?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/2e27cff7bd9c</guid>
            <category><![CDATA[langchain-vs-langgraph]]></category>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Tue, 17 Jun 2025 07:24:03 GMT</pubDate>
            <atom:updated>2025-06-17T07:24:03.027Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*DVTfbeVDA2q9d7FH" /><figcaption>Photo by <a href="https://unsplash.com/@siderius_creativ?utm_source=medium&amp;utm_medium=referral">Gerard Siderius</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Introduction:</h3><p>Building an LLM assistant often <strong>starts</strong> as a straight-line recipe: <em>retrieve</em> the right passages, <em>summarise</em> them, and present a crisp answer. That linear mindset is exactly what <strong>LangChain</strong> optimises for — every step runs once, top-to-bottom, inside a directed-acyclic-graph (DAG) chain.</p><p>Fast-forward a few sprints and the same product suddenly needs two cooperating agents, a memory that survives crashes, and a human reviewer in the loop. A plain chain can’t loop back or share state comfortably, which is where <strong>LangGraph</strong> steps in. By treating every step as a node on a <strong>graph</strong> — not a one-way DAG — it unlocks conditional branches, cycles, checkpoints, and truly persistent memory. solve the same task in both frameworks, and a quick verdict on when to choose each.</p><h3>Key design bullet points:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1019/1*nbAKVHb1isIYdrL32kZf9Q.png" /><figcaption>comparison table</figcaption></figure><p>Code showdown — “Why did we switch to LangGraph?”</p><ol><li>LangChain (sequential RAG)</li></ol><pre># pip install langchain-community langchain-openai faiss-cpu<br><br>from langchain.text_splitter import RecursiveCharacterTextSplitter<br>from langchain.embeddings         import OpenAIEmbeddings<br>from langchain.vectorstores       import FAISS<br>from langchain.docstore.document  import Document<br>from langchain.chat_models        import ChatOpenAI<br>from langchain.chains             import RetrievalQA<br><br># -- ingestion -------------------------------------------------<br>raw_docs = [Document(page_content=open(&quot;docs/article.txt&quot;).read())]<br>splits    = RecursiveCharacterTextSplitter().split_documents(raw_docs)<br><br># -- vector index ---------------------------------------------<br>vectordb  = FAISS.from_documents(<br>              splits,<br>              OpenAIEmbeddings(model=&quot;text-embedding-3-small&quot;)<br>           )<br><br># -- run chain -------------------------------------------------<br>llm  = ChatOpenAI(model_name=&quot;gpt-4o-mini&quot;)<br>rag  = RetrievalQA.from_chain_type(<br>          llm,<br>          chain_type=&quot;stuff&quot;,<br>          retriever=vectordb.as_retriever()<br>      )<br><br>print(rag(&quot;Why did we switch to LangGraph?&quot;)<br><br>Note:Everything executes once, in order; no state survives the call.</pre><p>2. LangGraph (cyclic, multi-agent, persistent)</p><pre># pip install langgraph langchain-openai<br><br>import langgraph<br>from langgraph.checkpoint import SqliteCheckpoint<br>from langgraph.graph.message import MessageGraph<br>from langchain.chat_models import ChatOpenAI<br><br># shared memory across the whole graph ------------------------<br>checkpointer = SqliteCheckpoint(&quot;agent_state.sqlite&quot;)<br>graph        = MessageGraph(checkpointer=checkpointer)<br><br># task nodes ---------------------------------------------------<br>def retrieve(state):<br>    docs = vectordb.similarity_search(state[&quot;question&quot;])<br>    return {&quot;docs&quot;: docs}<br><br>def draft(state):<br>    llm     = ChatOpenAI(model_name=&quot;gpt-4o-mini&quot;)<br>    context = &quot;\n&quot;.join(d.page_content for d in state[&quot;docs&quot;])<br>    answer  = llm.invoke(f&quot;{context}\n\nQ: {state[&#39;question&#39;]}&quot;)<br>    return {&quot;draft&quot;: answer}<br><br>def review(state):<br>    # human gatekeeper<br>    print(&quot;\n--- DRAFT ---\n&quot;, state[&quot;draft&quot;])<br>    approved = input(&quot;Approve? (y/n) &quot;)<br>    return {&quot;approved&quot;: approved.lower() == &quot;y&quot;}<br><br># wiring -------------------------------------------------------<br>graph.add_node(&quot;retrieve&quot;, retrieve)<br>graph.add_node(&quot;draft&quot;,    draft)<br>graph.add_node(&quot;review&quot;,   review)<br><br>graph.add_edge(&quot;retrieve&quot;, &quot;draft&quot;)<br><br># branch: if LLM mentions LangGraph, send to human reviewer<br>graph.add_conditional_edges(<br>    &quot;draft&quot;,<br>    lambda s: (&quot;review&quot; if &quot;LangGraph&quot; in s[&quot;draft&quot;].content else &quot;done&quot;)<br>)<br><br># loop: on rejection, go back to retrieval to try again<br>graph.add_conditional_edges(<br>    &quot;review&quot;,<br>    lambda s: (&quot;done&quot; if s[&quot;approved&quot;] else &quot;retrieve&quot;)<br>)<br><br>agent = graph.compile()<br><br>print(agent.invoke({&quot;question&quot;: &quot;Why did we switch to LangGraph?&quot;}))<br><br>Note:Notice how the same vectordb and the conversation history live in the checkpointed state, so you can crash, restart, or hand off to another process without losing context.<br><br></pre><h3>Conclusion:</h3><p><strong>Stick with LangChain</strong> : when your workflow is naturally linear — document chatbots, FAQ assistants, or anything that ends as soon as an answer is printed. Its rich catalog of loaders and vector-store integrations gets you from zero to demo in minutes.</p><p><strong>Reach for LangGraph</strong> once you need loops, branches, multiple agents, or memory that spans hours and restarts. Cyclic edges, checkpointing, and shared state turn “one-shot” chains into resilient, production-ready agents.</p><h3><strong>References:</strong></h3><p><a href="https://oxylabs.io/blog/langgraph-vs-langchain">LangChain vs. LangGraph: Comparing AI Agent Frameworks</a></p><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*in-_rTJtip1YqsZ4" /><figcaption>Photo by <a href="https://unsplash.com/@kevin_xue_pic?utm_source=medium&amp;utm_medium=referral">kevin Xue</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2e27cff7bd9c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“The Power of RAGAS: Automated Evaluation for Enhanced Content Generation.”]]></title>
            <link>https://medium.com/@karjalpp/the-power-of-ragas-automated-evaluation-for-enhanced-content-generation-0c39602c2d80?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/0c39602c2d80</guid>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Sat, 20 Apr 2024 15:55:48 GMT</pubDate>
            <atom:updated>2024-04-20T15:55:48.495Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xgcdBU2ds9r47eLJ" /><figcaption>Photo by <a href="https://unsplash.com/@sonance?utm_source=medium&amp;utm_medium=referral">Viktor Forgacs™️</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p><strong>Introduction</strong>:</p><p>RAGAS (Retrieval Augmented Generation Assessment) is introduced as a framework designed for the reference-free evaluation of Retrieval Augmented Generation (RAG) pipelines. These RAG systems consist of two main components: a retrieval module and a Language Model (LLM)-based generation module. They enable LLMs to interact with textual databases, mitigating the risk of generating false or misleading information (hallucinations).</p><p>Evaluating RAG architectures poses challenges due to multiple dimensions: the retrieval system’s capacity to pinpoint relevant and precise context passages, the LLM’s ability to effectively utilize such passages, and the overall quality of the generated content.</p><p>With RAGAS, we propose a suite of metrics capable of assessing these different dimensions without relying on human-provided ground truth annotations. This framework aims to significantly expedite the evaluation process of RAG architectures, a critical need given the rapid integration of Large Language Models (LLMs) in various applications.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oeEjAlsphXFVmsUx8eDXhg.png" /></figure><p>To evaluate the RAG pipeline, RAGAs expects the following information:</p><ul><li>question: The user query that is the input of the RAG pipeline. The input.</li><li>answer: The generated answer from the RAG pipeline. The output.</li><li>contexts: The contexts retrieved from the external knowledge source used to answer the question.</li><li>ground_truths: The ground truth answer to the question. This is the only human-annotated information. This information is only required for the metric context_recall (see <a href="https://towardsdatascience.com/evaluating-rag-applications-with-ragas-81d67b0ee31a#c52f">Evaluation Metrics</a>).</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/753/1*29LUAWUyghKctqMzn1LSgw.png" /></figure><p><strong>Evaluation Steps:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/644/1*xYlWyce5WlfT4RSP9N1GeA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w-LaCLVeuBbYvPTq7XOIDQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AJYCsTRUKJ9qCIiQAkAQ5A.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/787/1*a920IguBx7jkw_TcoBxfNQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/642/1*EUkMPEe9F6foMsGgg8dbxA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/839/1*o2yExt_jeihRnr7wRiKNhA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pKHO6A1eJ8YESDwV62uFpw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/754/1*I6Pd4vIzXzpfYB99hiQokw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4CLihRTheGtLvG5vJJzaZQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XUaHt4LHUutsDCmu1LEzfg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*euw8LA943cIA1rbX6l6Z-w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yFC_IPsPrt6f7W2pXhP2Lw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CJkEJCQ0D8mmNcd6kIlzYA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bPnog4mJMfMz2TL0lwSIcA.png" /><figcaption>data frame based on union.txt data example of evaluation matrix.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i1ch86MPPriQJTNN4DITYA.png" /><figcaption>data frame similarity matrix based on answer and ground truth.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7ZyRU0XTbO5BbaNFf1auqw.png" /><figcaption>data frame based on answer correctness to compare answers according to knowledge base.</figcaption></figure><p><strong>Summary:</strong></p><p>The article discusses the challenges in transitioning RAG (Retrieval-Augmented Generation) application to a production-ready state. It emphasizes the importance of evaluating the RAG pipeline’s performance using a validation dataset and metrics, which can be complex due to the multiple components involved.</p><p>To address this complexity, the article introduces the RAGAs evaluation framework, which proposes four key metrics — context_relevancy, context_recall, faithfulness, and answer_relevancy — to assess the RAG pipeline comprehensively. Additionally, RAGAs leverages Large Language Models (LLMs) for reference-free evaluation, aiming to mitigate the costs associated with creating high-quality validation datasets.</p><p>By employing the RAGAs framework, developers can better evaluate their RAG applications’ performance and optimize them using experimentation pipelines and tuning strategies provided in the article.</p><p>References: <a href="https://towardsdatascience.com/evaluating-rag-applications-with-ragas-81d67b0ee31a">Evaluating RAG Applications with RAGAs | by Leonie Monigatti | Towards Data Science</a></p><p><a href="https://docs.ragas.io/en/stable/">Introduction | Ragas</a></p><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Kl_oqg99nJCws2VT" /><figcaption>Photo by <a href="https://unsplash.com/@peet818?utm_source=medium&amp;utm_medium=referral">Pete Pedroza</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0c39602c2d80" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“Future Unleashed: Sora — OpenAI’s Text-to-Video Model and Its Revolutionary Applications”.]]></title>
            <link>https://medium.com/@karjalpp/future-unleashed-sora-openais-text-to-video-model-and-its-revolutionary-applications-f9b9cedd46fb?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/f9b9cedd46fb</guid>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Tue, 20 Feb 2024 13:53:57 GMT</pubDate>
            <atom:updated>2024-02-20T13:53:57.462Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>“Future Unleashed: Sora — OpenAI’s Text-to-Video Model and Its Revolutionary Applications”.</strong></p><figure><img alt="Sora (openai.com)" src="https://cdn-images-1.medium.com/max/1024/1*HBQjwgrxu5DtBLjBic-WuA.png" /></figure><p>The primary objective of Sora, <strong>OpenAI’s text-to-image converter,</strong> is to bridge the gap between textual descriptions and visual representations by generating realistic and contextually relevant images from given textual input. This technology aims to enable various applications such as creative content generation, visual storytelling, augmented reality, virtual environments, and more, by automating the process of image creation based on textual descriptions. Sora seeks to push the boundaries of artificial intelligence and enhance human-computer interaction by providing a powerful tool for transforming ideas and concepts into visually compelling imagery.</p><p>look at the image /video Tour of an art gallery it&#39;s <strong>Amazing!!!</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vOK0aB_YaRz6gF6WHuEGyw.png" /></figure><p><a href="https://cdn.openai.com/sora/videos/art-museum.mp4">cdn.openai.com/sora/videos/art-museum.mp4</a></p><p>While OpenAI’s Sora is currently not publicly available, information about its underlying technology is limited. However, based on the glimpses and explanations provided by OpenAI, here’s what we know:</p><p><strong>Text Processing:</strong></p><ul><li>Likely uses large language models (LLMs) similar to OpenAI’s GPT-3, which can understand complex language and generate creative text formats.</li><li>These LLMs might be further fine-tuned to analyze and comprehend specific video-related prompts and scripts.</li></ul><p><strong>Video Generation:</strong></p><ul><li>The core is likely a generative adversarial network (GAN) trained on massive datasets of images and videos.</li><li>One network (generator) creates video frames based on the text input and understanding from the LLM.</li><li>Another network (discriminator) evaluates the generated frames for realism and consistency with the input.</li><li>Through constant competition and iteration, the generator improves its ability to create realistic and coherent video sequences.</li></ul><p><strong>Additional Techniques:</strong></p><ul><li>Sora might employ techniques like optical flow for smooth scene transitions and camera movements.</li><li>Animation principles could be incorporated for character movement and expression.</li><li>Style transfer models might be used to apply specific artistic styles to the generated video.</li></ul><p>Potential applications of Sora in near future:</p><ol><li>Data Visualization: Sora’s ability to translate textual descriptions into visual representations can significantly enhance data exploration and communication.</li><li>Automated Report Generation: By automating the creation of visual elements, Sora can streamline the report generation process, allowing to focus more on analysis and insights.</li><li>Exploratory Data Analysis (EDA): Sora’s visualizations can aid in understanding complex datasets, identifying patterns, and detecting anomalies, thereby facilitating more informed decision-making.</li><li>Prototype and Mock-Up Creation: Sora’s capacity to generate realistic images from text descriptions enables rapid prototyping and visualization of data-driven concepts, facilitating the design process.</li><li>Data Augmentation: Sora’s ability to generate synthetic images based on text descriptions can enrich datasets, potentially improving the performance and generalization of machine learning and deep learning models.</li></ol><h3>Core Capabilities:</h3><ul><li>Sora is a powerful diffusion model developed by OpenAI, designed to generate videos from textual descriptions or existing images/videos.</li><li>It operates by gradually removing noise from an initial frame, allowing it to create seamless and coherent video sequences. Leveraging a transformer architecture, similar to GPT models, Sora achieves superior scaling performance and can handle a wide range of visual data representations. Drawing from previous research in models like DALL·E and GPT, Sora incorporates techniques such as recaptioning to generate descriptive captions, ensuring fidelity to user instructions.</li><li>Sora’s versatility extends to animating still images, extending videos, and filling in missing frames. Beyond its immediate applications, Sora represents a foundational step toward achieving Artificial General Intelligence (AGI) by simulating and understanding the complexities of the real world.</li></ul><p><strong>Uncertainties:</strong></p><ul><li>The specific details of the AI models and training data used are not publicly known.</li><li>The extent to which human involvement is used in the editing or refinement process is unclear.</li><li>Ethical considerations about potential misuse and generation of misinformation remain a concern.</li></ul><p>References: <a href="https://openai.com/sora">Sora (openai.com)</a></p><p><strong>It’s important to remember that Sora is still under development, and the technology is constantly evolving. As more information becomes available, we’ll gain a deeper understanding of its inner workings and capabilities.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yTAfMy9Y8eTg4PRkQdDeUA.png" /></figure><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>. <em>I promise it will be unique!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f9b9cedd46fb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“Gemini Pro in Action: Case Studies and Success Stories”]]></title>
            <link>https://medium.com/datadreamers/gemini-pro-in-action-case-studies-and-success-stories-b5cd607f7192?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/b5cd607f7192</guid>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Wed, 27 Dec 2023 15:58:29 GMT</pubDate>
            <atom:updated>2024-01-12T19:10:47.950Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/902/1*18o4dh19H--LivGAjDgRGQ.png" /></figure><p><strong>Abstract</strong>: Let&#39;s start with Navigating the “Gemini Pro Multimodal Landscape” for different applications.</p><p>“Gemini “Multimodal AI application is a system that can analyze both text and images in tandem. This could be useful in tasks such as image captioning, where the model generates a textual description based on the content of an image.</p><p><strong>The Gemini Pro model </strong>is capable of handling text generation and multi-turn chat conversations. Just like any other Large Language Model, the Gemini Pro can do in-context learning and can handle zero-shot, one-shot, and few-shot prompting techniques. According to the official documentation, the input token limit for the Gemini Pro LLM is 30,720 tokens and the output limit is 2048 tokens. And for the free API, Google allows up to <strong>60 requests per minute.</strong></p><p><strong>The Gemini Pro Vision</strong> is a multimodal built to work on tasks that require Large Language Models to understand Images. The model takes both the Image and Text as input and generates text as output. The Gemini Pro Vision can take up to 12,288 tokens through the input and can output up to a maximum of 4096 tokens. Like the Gemini Pro, this model can even handle different prompt strategies like zero-shot, one-shot, and few-shot.</p><h3>Gemini API: QuickStart with Python</h3><ol><li>Set up your development environment and API access to use Gemini.</li><li>Generate text responses from text inputs.</li><li>Generate text responses from multimodal inputs (text and images).</li><li>Use Gemini for multi-turn conversations (chat).</li><li>Use embeddings for large language models.</li></ol><h3>Prerequisites:</h3><ul><li>Python 3.9+ or 3.10 for better results.</li></ul><h3>Setup:</h3><h3>Install the Python SDK:</h3><p>! pip install -q -U google-generativeai</p><h3>Import packages:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x7TMUHGoheXJahhUsdulTA.png" /></figure><h3>Setup your API key:</h3><p><a href="https://makersuite.google.com/app/apikey">Get API key | Google AI Studio</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/869/1*7NOMWi1OAm-MGp-iTWaY3w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/951/1*N8Xhgrm7BLnQjAuyUU40KQ.png" /></figure><h3>List models:</h3><ul><li>gemini-pro: optimized for text-only prompts.</li><li>gemini-pro-vision: optimized for text-and-images prompts.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/741/1*hZvtbL_JMCOVE5PP65yQLQ.png" /></figure><h3>Generate text from text inputs:</h3><p>model = genai.GenerativeModel(‘gemini-pro’)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/880/1*oyz1Qu9U7M3Sp7PILcyI4A.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RNq3EQPVmG5Y4DafJkUsNg.png" /></figure><p>Its good feature gemini have:Gemini can generate multiple possible responses for a single prompt. These possible responses are called candidates, and you can review them to select the most suitable one as the response.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wGC5b91rKJNArZSWRMn8pA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZeFWdsQGS1mJuwPycd1FRQ.png" /><figcaption>chunk response</figcaption></figure><h3>Generate text from image and text inputs:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GZ3z1DMZV_0tGf8vWlR28A.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/978/1*6Hel77q0vnpGKCpNI-G6ZQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*75D8DMBOPHBFmpe1t3SISQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/873/1*Bo77DGBjHqACunLJ9wihSw.png" /></figure><p>Gemini enables you to have freeform conversations across <strong>multiple turns.</strong> The ChatSession class simplifies the process by managing the state of the conversation, so unlike with generate_content, you do not have to store the conversation history as a list.</p><h3>Use embeddings:</h3><p><a href="https://developers.google.com/machine-learning/glossary#embedding-vector">Embedding</a> is a technique used to represent information as a list of floating-point numbers in an array. With Gemini, you can represent text (words, sentences, and blocks of text) in a vectorized form, making it easier to compare and contrast embeddings. For example, two texts that share a similar subject matter or sentiment should have similar embeddings, which can be identified through mathematical comparison techniques such as cosine similarity. For more on how and why you should use embeddings, refer to the <a href="https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fai.google.dev%2Fdocs%2Fembeddings_guide">Embeddings guide</a>.</p><p>Use the embed_content method to generate embeddings. The method handles embedding for the following tasks (task_type):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/970/1*dzLADCo3kchJ3D_3Zs_mXw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/856/1*NRn9r9GtiK-7-3ktLH3_vg.png" /></figure><p><strong>What’s next:</strong></p><ul><li>Prompt design is the process of creating prompts that elicit the desired response from language models. Writing well structured prompts is an essential part of ensuring accurate, high quality responses from a language model. Learn about best practices for <a href="https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fai.google.dev%2Fdocs%2Fprompt_best_practices">prompt writing</a>.</li><li>Gemini offers several model variations to meet the needs of different use cases, such as input types and complexity, implementations for chat or other dialog language tasks, and size constraints. Learn about the available <a href="https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fai.google.dev%2Fmodels%2Fgemini">Gemini models</a>.</li><li>Gemini offers options for requesting <a href="https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fai.google.dev%2Fdocs%2Fincrease_quota">rate limit increases</a>. The rate limit for Gemini-Pro models is 60 requests per minute (RPM).</li></ul><p>Applications of Multimodal in future:</p><ol><li><strong>Healthcare:</strong></li></ol><p>i)Medical Imaging Analysis: Multimodal models can simultaneously analyze medical images (such as MRIs or CT scans) and textual patient records to provide more comprehensive diagnostics.</p><p>ii)Health Monitoring: Integrating data from wearable devices, like combining sensor data with patient-reported information, can enhance health monitoring systems.</p><p><strong>2.Autonomous Vehicles:</strong></p><p>i) Sensor Fusion: Multimodal models integrate data from various sensors, including cameras, LiDAR, and radar, to improve perception and decision-making in autonomous vehicles.</p><p>ii) Natural Language Interaction: Combining visual and auditory inputs with natural language processing enables better communication between vehicles and users.</p><p>3.<strong>Entertainment:</strong></p><p>i)Content Creation: Multimodal models can assist in generating diverse content, such as combining text and images for storytelling or creating realistic dialogue for virtual characters.</p><p>ii)Emotion Recognition: Analyzing facial expressions, voice tone, and textual sentiment together allows for more accurate emotion recognition in virtual environments.</p><p><strong>4.Education:</strong></p><p>i)Personalized Learning: Multimodal models can assess a student’s performance using data from various sources, including text-based assessments, speech interactions, and visual cues, enabling personalized learning experiences.</p><p>ii)Language Learning: Combining text, audio, and visual content can enhance language learning applications by providing a more immersive experience.</p><p><strong>5.E-commerce</strong>:</p><p>i)Visual Search: Multimodal models power visual search engines, allowing users to search for products using images rather than text.</p><p>ii)Customer Interaction: Integrating text and voice data in customer interactions enables more natural and context-aware virtual assistants or chatbots.</p><p>6. <strong>Security and Surveillance:</strong></p><p>i) Behavior Analysis: Combining video feeds with audio and text data can improve the analysis of human behavior in security and surveillance applications.</p><p>ii) Anomaly Detection: Multimodal models can identify anomalies by considering data from different sources, such as combining visual and sensor data in critical infrastructure monitoring.</p><p><strong>7.Finance:</strong></p><p>i)Fraud Detection: Multimodal models can improve fraud detection by combining transaction data, textual information, and visual cues.</p><p>ii)Sentiment Analysis: Analyzing financial news articles along with market data helps in better sentiment analysis for investment decisions.</p><p>References: 1. <a href="https://www.analyticsvidhya.com/blog/2023/12/building-a-multimodal-chatbot-with-gemini-and-gradio/">Building a MultiModal Chatbot with Gemini and Gradio — Analytics Vidhya</a></p><p>2.<a href="https://www.youtube.com/watch?v=W0rNUAfqGFc">Google Gemini Pro LLM Model Free API Demo With Code- Is It Better Than OpenAI GPT’s? (youtube.com)</a></p><p><a href="https://github.com/google/generative-ai-docs/blob/main/site/en/tutorials/python_quickstart.ipynb">3.generative-ai-docs/site/en/tutorials/python_quickstart.ipynb at main · google/generative-ai-docs · GitHub</a></p><p><strong>Conclusion:</strong> Gemini Pro Multimodal emerges as a pioneering platform that transcends traditional boundaries, offering a holistic and dynamic approach to creative expression.</p><p>From seamlessly fusing text and images to exploring the auditory dimensions through audio integration, Gemini Pro proves to be a versatile companion for those seeking to amplify their creative endeavors. Its user-friendly interface and progressive features provide a welcoming entry point for beginners, while its advanced capabilities cater to the needs of seasoned creators, making it a truly inclusive platform.</p><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>. <em>I promise it will be unique!</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*MZxQMR5irr1GVhC7" /><figcaption>Photo by <a href="https://unsplash.com/@calebchen?utm_source=medium&amp;utm_medium=referral">Caleb Chen</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b5cd607f7192" width="1" height="1" alt=""><hr><p><a href="https://medium.com/datadreamers/gemini-pro-in-action-case-studies-and-success-stories-b5cd607f7192">“Gemini Pro in Action: Case Studies and Success Stories”</a> was originally published in <a href="https://medium.com/datadreamers">DataDreamers</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“Einstein GPT Unleashed: Revolutionizing Service Reply Dynamics through Summarization and…]]></title>
            <link>https://medium.com/datadreamers/einstein-gpt-unleashed-revolutionizing-service-reply-dynamics-through-summarization-and-grounding-c480188ee392?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/c480188ee392</guid>
            <category><![CDATA[salesforce-einstein-gpt]]></category>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Sat, 23 Dec 2023 02:50:59 GMT</pubDate>
            <atom:updated>2024-01-12T19:10:27.809Z</atom:updated>
            <content:encoded><![CDATA[<h4>“Einstein GPT Unleashed: Revolutionizing Service Reply Dynamics through Summarization and Grounding Brilliance”</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lKC-1G3abf8svEFSCaOMiA.png" /></figure><p><strong>Abstract:</strong> Step into the realm of artificial intelligence as we explore the transformative prowess of Einstein GPT, a fusion of genius inspired by the legendary Albert Einstein and the cutting-edge technology of OpenAI’s GPT architecture.</p><p>With a keen focus on summarization and grounding, we unravel the capabilities of Einstein GPT in simplifying complex information and refining responses to an unprecedented level of accuracy.</p><p>Join with me on this enlightening journey as we unravel the potential of Einstein GPT to redefine service reply to dynamics, setting the stage for a new era in efficient and intelligent communication.</p><p>Saves agents time with AI-generated case summaries. Based on a conversation between an agent and customer, Einstein predicts and fills a summary, issue, and resolution. Agents can then review, edit, and save these summaries. With Conversation Catch-Up, Einstein also shows mid-conversation summaries to agents and supervisors when they accept or monitor an ongoing conversation.</p><p>Requirement: SDO (Simple Demo Org) allows for partners to access and demo Salesforce functionality for both individual education as well as for light demos for prospects and customers. The SDO is freely available to partners via Partner Learning Camp (PLC).</p><p>Agent-Customer Interaction Demo: Elevating Service Reply Work with <strong>Einstein GPT work Summariser.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mMFG-8DrVDpaTaEEJ08bKg.png" /><figcaption>Agent-Customer Interaction Demo for Browser Issue Obtained.</figcaption></figure><h4><strong>Use Case: 1 Utilizing Einstein GPT Service reply for Summarization:</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/882/1*0rNr4yeSOo8qXMqJ7tqXYQ.png" /><figcaption>Saves agents time with AI-generated case summaries. Based on a conversation between an agent and customer, Einstein predicts and fills a summary, issue, and resolution. Agents can then review, edit, and save these summaries.</figcaption></figure><p>Bring Agents and Supervisors Up to Speed with Mid-Conversation Summaries. When an agent or supervisor joins an active voice call or messaging session, give them context quickly with a real-time AI-generated summary of the conversation.</p><p><strong>Use Case: 2 Service Reply with Grounding: Elevating Customer Support with Einstein GPT</strong></p><p>Use Einstein AI to draft responses using a defined data source. Grounding indexes the objects and fields so that Einstein knows which information to base recommendations on. With grounding, your unique knowledge articles and case history add context and personalization to customer communications.</p><p>In the dynamic world of customer support, clarity and precision in service replies are paramount. Einstein GPT emerges as a game-changer, seamlessly integrating into customer-agent interactions to provide not only succinct summaries but also contextually grounded responses. Let’s explore a use case to understand how Einstein GPT transforms a service reply with its advanced grounding capabilities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/950/1*kKTY8lniyObXDKvX8fAYPw.png" /><figcaption>Grounding with Knowledge articles</figcaption></figure><p><strong>Future Scope: </strong>The future scope of integrating Einstein GPT into service reply work is vast and promising. As technology continues to evolve, the potential advancements and applications of this innovative AI model are expected to shape the landscape of customer support and communication. Here are several aspects that represent the future scope of Einstein GPT in service reply scenarios we are waiting:</p><ol><li><strong>Enhanced Personalization:</strong> Einstein GPT has the potential to analyze historical interactions and customer preferences, enabling even more personalized and context-aware service replies. This could result in tailored solutions that anticipate customer needs, fostering a more engaging and satisfying experience.</li><li><strong>Expanded Industry Applications: </strong>The principles of Einstein GPT can be applied across various industries beyond customer support, such as legal, healthcare, and education. Its ability to understand and generate human-like responses makes it adaptable to diverse professional contexts.</li><li><strong>Continuous Learning and Adaptability: Future</strong> versions of Einstein GPT may incorporate advanced learning mechanisms, allowing the model to adapt and evolve based on changing customer preferences, industry trends, and evolving service requirements.</li><li><strong>Ethical Considerations and Bias Mitigation: The</strong> future of AI involves addressing ethical concerns and ensuring fairness in interactions. Future versions of Einstein GPT may focus on refining algorithms to mitigate biases and prioritize ethical considerations in generating service replies.</li><li><strong>Feedback Loops for Model Improvement: Continuous</strong> feedback from users and support agents could contribute to refining and enhancing the performance of Einstein GPT. This iterative process could lead to increasingly accurate and contextually aware service replies.</li></ol><p><strong>Conclusion:</strong> In this use cases, Einstein GPT not only provides a concise summary of the customer’s issue but also grounds the response in specific, actionable steps. This grounded approach ensures that the customer not only understands the resolution but can also follow a structured path to address the issue.</p><p>Einstein GPT’s unique ability to combine summarization and grounding transforms service replies into informative, personalized conversations, enhancing customer satisfaction and support efficiency. As we witness this integration, it becomes evident that Einstein GPT is not just an AI model; it’s a catalyst for elevating the customer support experience to unprecedented heights.</p><p><strong>References:</strong><a href="https://blog.varianceinfotech.com/salesforce-einstein-gpt-how-does-it-work/">1 new message (varianceinfotech.com)</a><a href="https://quip.com/m3gMAUjabbW3">AI &amp; Einstein | Partner Pocket Guide (fka Einstein GPT Pocket Guide) — Quip</a><a href="https://trailhead.salesforce.com/content/learn/modules/einstein-gpt-quick-look/get-started-with-einstein-gpt">Get Started with Einstein Generative AI Unit | Salesforce Trailhead</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Vvt7z00-LOfInyeClQBjWA.png" /></figure><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>. <em>I promise it will be unique!</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*J1CSYQAE9r-ZYhTn" /><figcaption>Photo by <a href="https://unsplash.com/@alexas_fotos?utm_source=medium&amp;utm_medium=referral">Alexas_Fotos</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c480188ee392" width="1" height="1" alt=""><hr><p><a href="https://medium.com/datadreamers/einstein-gpt-unleashed-revolutionizing-service-reply-dynamics-through-summarization-and-grounding-c480188ee392">“Einstein GPT Unleashed: Revolutionizing Service Reply Dynamics through Summarization and…</a> was originally published in <a href="https://medium.com/datadreamers">DataDreamers</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“Exploring the Road Ahead: The Future of Data Science and Its Intersection with Gen AI.]]></title>
            <link>https://medium.com/datadreamers/exploring-the-road-ahead-the-future-of-data-science-and-its-intersection-with-gen-ai-65704b4d7f6e?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/65704b4d7f6e</guid>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Wed, 11 Oct 2023 15:24:55 GMT</pubDate>
            <atom:updated>2023-10-11T15:58:44.705Z</atom:updated>
            <content:encoded><![CDATA[<p>“<strong>Exploring the Road Ahead: The Future of Data Science and Its Intersection with Gen AI.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*adUQcr56ewlCaqGW" /><figcaption>Photo by <a href="https://unsplash.com/@siderius_creativ?utm_source=medium&amp;utm_medium=referral">Gerard Siderius</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>The convergence of data science and Gen AI signifies a significant leap forward in our ability to extract meaning from the ever-expanding sea of data. In this article, will delve into this exciting partnership, exploring how data scientists are poised to shape the future of their field and, in turn, transform industries, advance research, and drive innovation in ways we have never before imagined.</p><p>a new frontier is emerging — the realm of “Gen AI.” Data scientists, the modern-day alchemists of insights and knowledge, are at the forefront of this transformative journey. As they harness the power of Gen AI, a groundbreaking synergy is taking shape, promising to redefine the very essence of data science.</p><p>The future scope for data scientists in the context of the intersection with artificial general intelligence (Gen AI) is incredibly promising. Here are some key areas where data scientists can expect to play a crucial role:</p><p>The future scope for data scientists in the context of the intersection with artificial general intelligence (Gen AI) is incredibly promising.</p><p><strong>Here are</strong> <strong>some key areas where data scientists can expect to play a crucial role:</strong></p><ol><li><strong>Advanced AI Model Development</strong>: As Gen AI continues to evolve, data scientists will be at the forefront of developing and fine-tuning the AI models that underpin it. They will work on improving the performance, efficiency, and accuracy of these models, enabling them to adapt to a wider range of tasks and domains.</li><li><strong>Data Quality and Curation:</strong> High-quality data is the lifeblood of AI, including Gen AI. Data scientists will be responsible for ensuring data quality, data cleaning, and curation to feed Gen AI models with the best possible data. This includes addressing bias, noise, and ensuring data privacy.</li><li><strong>Interpretable and Ethical AI:</strong> Data scientists will be vital in making AI and Gen AI models more interpretable and transparent. They will work on developing techniques that help explain AI decisions and ensure ethical considerations are embedded into the models.</li><li><strong>Data-Driven Decision-Making:</strong> Data scientists will continue to support organizations in making data-driven decisions. Gen AI can generate insights, but data scientists can help interpret and apply these insights effectively in a business context.</li><li><strong>AI Governance and Regulation:</strong> With the rise of Gen AI, the need for governance and regulation will become more pressing. Data scientists will be involved in developing and implementing policies that ensure the responsible and ethical use of AI.</li><li><strong>Cross-Disciplinary Collaboration: </strong>The future of data science and Gen AI will involve more collaboration with experts from various fields. Data scientists will need to work closely with domain specialists to ensure AI models are tailored to specific applications.</li><li><strong>AI-Driven Research and Innovation:</strong> Data scientists will be integral in conducting research and innovation using AI, as Gen AI can accelerate scientific discovery in various fields, from healthcare to material science.</li><li><strong>Data Security and Privacy:</strong> As data becomes more valuable and privacy concerns increase, data scientists will be tasked with protecting sensitive information and developing techniques to maintain data security</li><li><strong>Education and Skill Development:</strong> Data scientists will also play a role in educating and mentoring the next generation of AI professionals, sharing their expertise in data science, AI, and Gen AI.</li><li><strong>Personalized AI Solutions: </strong>Data scientists will help in creating highly personalized AI solutions. Gen AI can adapt to individual preferences and needs, and data scientists will fine-tune these models to deliver optimal user experiences.</li></ol><p><strong>Challenges with Gen AI:</strong></p><p>Data scientists working with Gen AI in the future will need to be versatile, agile, and well-versed in a wide range of skills, from data management and modeling to ethics and regulation. They will play a crucial role in ensuring the responsible and ethical development and deployment of Gen AI systems.</p><ol><li><strong>Ethical and Regulatory Challenges:</strong> With the increasing adoption of Gen AI, there will be more concerns about ethical issues like privacy, bias, transparency, and accountability. Data scientists will need to navigate a complex landscape of regulations and standards to ensure ethical AI development and deployment.</li><li><strong>Data Management: </strong>Handling vast amounts of data will become even more challenging. Data scientists will need to develop strategies for efficiently collecting, storing, and managing large datasets. They’ll also need to ensure data quality and security.</li><li><strong>Interoperability and Integration: </strong>As Gen AI systems become more diverse and specialized, integrating them into existing systems and workflows will require expertise in ensuring compatibility and functionality.</li><li><strong>Model Complexity:</strong> Gen AI models are expected to be highly complex and difficult to understand fully. Data scientists will need to develop techniques for model interpretability and explainability to build trust and compliance with stakeholders and regulators.</li><li><strong>Resource Intensity:</strong> Developing and training Gen AI models can be computationally expensive. Data scientists will need to find ways to make these processes more efficient, sustainable, and cost-effective.</li><li><strong>Skills and Talent Gap: </strong>The demand for data scientists with expertise in Gen AI will likely outstrip the supply. Preparing the workforce and upskilling data scientists to work with highly advanced AI systems will be a continuous challenge.</li><li><strong>Data Privacy and Security:</strong> As Gen AI becomes more integrated into our daily lives, there will be an increased risk of data breaches and cyberattacks. Data scientists must develop advanced techniques to protect data and ensure the security of AI systems.</li><li><strong>Bias Mitigation:</strong> Gen AI, like current AI, will inherit biases from the data it’s trained on. Data scientists will need to develop more robust techniques for identifying and mitigating these biases to ensure fair and equitable AI systems.</li><li><strong>Regulation and Compliance: </strong>As AI regulations evolve, data scientists will need to keep up with new requirements and ensure that AI systems are compliant. This includes data protection, transparency, and accountability regulations.</li><li><strong>AI Governance: </strong>Ensuring responsible AI governance will be a critical challenge. Data scientists will play a role in defining the governance structures that oversee AI development, use, and accountability.</li><li><strong>Sustainability: </strong>The environmental impact of large-scale AI and Gen AI models is a growing concern. Data scientists will need to work on developing energy-efficient AI algorithms and infrastructure.</li><li><strong>Unforeseen Ethical Dilemmas:</strong> Gen AI could lead to unforeseen ethical challenges. Data scientists will need to be adaptable and ready to address new ethical dilemmas as they arise.</li><li><strong>Cross-Disciplinary Collaboration: </strong>Collaboration with experts from various domains, such as ethics, law, sociology, and psychology, will be essential to address the multifaceted challenges associated with Gen AI.</li></ol><p><strong>Conclusion:</strong> the future of data science within the context of Gen AI is exciting. Data scientists will continue to be critical in building, fine-tuning, and governing AI models, ensuring their ethical and responsible use, and collaborating with other experts across various domains to leverage the full potential of Gen AI in solving complex problems and driving innovation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*QgTsktQTW2pc_bpA" /><figcaption>Photo by <a href="https://unsplash.com/@calebchen?utm_source=medium&amp;utm_medium=referral">Caleb Chen</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>. <em>I promise it will be unique!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=65704b4d7f6e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/datadreamers/exploring-the-road-ahead-the-future-of-data-science-and-its-intersection-with-gen-ai-65704b4d7f6e">“Exploring the Road Ahead: The Future of Data Science and Its Intersection with Gen AI.</a> was originally published in <a href="https://medium.com/datadreamers">DataDreamers</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AWS Serverless back to Monolith!]]></title>
            <link>https://medium.com/datadreamers/aws-serverless-back-to-monolith-308e74863db8?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/308e74863db8</guid>
            <category><![CDATA[api]]></category>
            <category><![CDATA[serverless-architecture]]></category>
            <category><![CDATA[monolithic-architecture]]></category>
            <category><![CDATA[api-development]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Thu, 01 Jun 2023 00:59:19 GMT</pubDate>
            <atom:updated>2023-06-01T00:59:19.709Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*FrOwM3anX3gC26vN" /></figure><p><strong>Objective</strong>: Transitioning from a <strong>serverless architecture</strong> to a M<strong>onolithic</strong> <strong>architecture</strong> on AWS is to simplify the development process and reduce the time and cost of maintenance. With a monolithic architecture, all components of an application are located in a single codebase, making it easier to manage and maintain. Additionally, it can reduce the complexity of the application and reduce the number of moving parts, making it more cost-effective and efficient.</p><p>Once upon a time, there was a business that relied on a monolith system to run its operations. The system was complex and expensive to maintain, and updates were hard to come by. <strong>The business was in need of a solution that would allow them to scale and keep up with the ever-changing technology landscape.</strong></p><p>Enter AWS serverless architecture. With serverless, the business was able to quickly spin up resources on demand and focus on building features and services rather than managing servers. <strong>The business could also reduce costs since they were only paying for the resources as they were needed instead of having to pay for a perpetually running infrastructure.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/978/1*40q5__oP9DzJj_Ojo5-ZrQ.png" /><figcaption>Fig:1 AWS serverless Architecture</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/682/1*NOVWeJxjerpN1OMQuh7MmA.png" /><figcaption>Fig:1 AWS Lambda serverless Architecture</figcaption></figure><p><strong>The business soon realized the benefits of serverless and decided to transition their existing monolith application to a serverless architecture. The transition was relatively smooth since the existing code was largely compatible with the serverless framework.</strong></p><p>The business was now able to take advantage of features <strong>such as auto-scaling, event-driven execution, and pay-per-use billing. This allowed them to focus on developing features and services that improved the customer experience</strong>. They could also utilize the serverless framework to deploy code faster and with fewer errors.</p><p>The business was able to reduce their costs and improve their application performance with the transition from monolith to serverless architecture. They now have the agility and scalability to keep up with the rapidly changing technology landscape.</p><p>In theory, the use of serverless would allow the team to scale each service independently. It turned out, however, that at least for how the team implemented the components, they hit a hard scaling limit at only 5% of the expected load. <strong>The costs of scaling up to monitor thousands of video streams would also be unduly expensive, due to the need to send data across multiple components.</strong></p><p>Initially, the team tried to optimize individual components, but this did not bring about significant improvements. <strong>So, the team moved all the components into a single process, hosting them on Amazon Elastic Compute Cloud (Amazon EC2) and Amazon Elastic Container Service (Amazon ECS).</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/1*o4obYhJ_cVr38CA5CB97sA.png" /><figcaption>fig:3 Amazon (EC2) &amp; Amazon (ECS)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B5QupW8JmuJN-5GVi-T2ZA.png" /><figcaption>fig:4 AWS monolithic architecture</figcaption></figure><p>The need for a monolithic AWS architecture compared to a serverless architecture depends on the<strong> needs of your project</strong>.</p><p><strong>Monolithic architectures are more suitable for applications that require complex processes and need to scale up quickly.</strong> <strong>They are also better for applications that have heavy traffic</strong>.</p><p>On the other hand, serverless architectures are better for applications that <strong>require more flexibility and are better for applications that don’t need to scale rapidly.</strong> Serverless architectures also <strong>usually cost less.</strong></p><p><strong>Benefits Of AWS serverless to Monolith:</strong></p><ol><li>AWS serverless architecture include scalability, cost efficiency, and automated management.</li><li>With serverless architecture, you don’t have to worry about managing and maintaining servers, which can lead to cost savings.</li><li>Additionally, it is easy to scale up resources when needed. The cons of serverless architecture include lack of control, complexity, and potential vendor lock-in.</li><li>You may have limited control over the underlying infrastructure and services, and the overall setup can be complex.</li><li>Lastly, depending on the service, you may be locked into a particular vendor.</li></ol><p><strong>Cons of AWS serverless to Monolith Architecture:</strong></p><ol><li>main downside to a serverless architecture over a monolith architecture is that it can be difficult to debug and troubleshoot issues.</li><li>Serverless architectures lack the ability to inspect server processes or logs, making it difficult to determine exactly what is causing an issue.</li><li>Additionally, serverless architectures can be more expensive to maintain due to the additional cost of cloud services.</li><li>Finally, serverless architectures require a greater level of expertise to fully utilize, making it more difficult for less experienced developers to understand and use.</li></ol><p>Challenges: <br> 1. <strong>Increased cost:</strong> While transitioning from serverless to monolith, the cost of maintaining the servers and the software will increase due to the need to run the application on a larger and more powerful server. <br> 2<strong>. Complexity of managing the infrastructure:</strong> Moving from serverless to monolith will require a more complex setup of servers, databases, and software, all of which need to be managed in order to keep the application running. <br> 3.<strong> Difficulty of debugging:</strong> Debugging an application running in a monolith architecture can be more difficult than debugging a serverless application, as it will require a deeper understanding of the underlying infrastructure and how it affects the application. <br> 4. <strong>Security risks:</strong> As the application will be running on a single server, any security breaches are more likely to have a greater impact on the application. <br> 5. <strong>Management overhead:</strong> With a monolith architecture, there will be more effort needed to manage the application and ensure it is running optimally.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jQ0mkU7K-FdcO22L" /><figcaption>Photo by <a href="https://unsplash.com/@grstocks?utm_source=medium&amp;utm_medium=referral">GR Stocks</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Thanks for Reading! …</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>. <em>I promise it will be unique!</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2PPYbDrrehjjAaaT" /><figcaption>Photo by <a href="https://unsplash.com/@alexas_fotos?utm_source=medium&amp;utm_medium=referral">Alexas_Fotos</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=308e74863db8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/datadreamers/aws-serverless-back-to-monolith-308e74863db8">AWS Serverless back to Monolith!</a> was originally published in <a href="https://medium.com/datadreamers">DataDreamers</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Revolutionizing Customer Service: How Intelligent Luis Chatbot Integration with ServiceNow is…]]></title>
            <link>https://medium.com/datadreamers/intelligent-luis-based-chatbot-integration-with-service-now-1fe82b51e035?source=rss-3d3ccfe5cc64------2</link>
            <guid isPermaLink="false">https://medium.com/p/1fe82b51e035</guid>
            <category><![CDATA[servicenow]]></category>
            <category><![CDATA[cloud-transformation]]></category>
            <category><![CDATA[intelligent-chatbot]]></category>
            <category><![CDATA[microsoft]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[KARJAL PRIYANKA]]></dc:creator>
            <pubDate>Thu, 20 Apr 2023 06:03:19 GMT</pubDate>
            <atom:updated>2023-04-28T02:13:23.880Z</atom:updated>
            <content:encoded><![CDATA[<h3>Revolutionizing Customer Service: How Intelligent Luis Chatbot Integration with ServiceNow is Streamlining Support Processes</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XGNPQaQo5A9tehU6" /><figcaption>Photo by <a href="https://unsplash.com/@santesson89?utm_source=medium&amp;utm_medium=referral">Andrea De Santis</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p><strong>Abstract</strong>: The objective of this project is to create an integration between Microsoft LUIS and Service Now. Microsoft LUIS is a AI natural language processing service that allows developers to create apps that can understand user input, and Service Now is a service desk application for managing IT operations. This integration will enable users to submit service requests through natural language, allowing for a more convenient, intuitive way for users to interact with Service Now. The integration will allow for natural language requests to be converted into Service Now requests, allowing for greater automation of the service desk operations.</p><p>Language Understanding (LUIS) is a cloud-based conversational AI service that applies custom machine-learning intelligence to a user’s conversational, natural language text to predict overall meaning, and pull out relevant, detailed information. LUIS provides access through its <a href="https://www.luis.ai/">custom portal</a>, <a href="https://go.microsoft.com/fwlink/?linkid=2092356">APIs</a> and <a href="https://learn.microsoft.com/en-us/azure/cognitive-services/luis/client-libraries-rest-api">SDK client libraries</a>.</p><p>Luis-based applications can support generic questions by leveraging the power of natural language understanding (NLU) technology to interpret user input and map it to the appropriate action. Additionally, many applications can allow users to upload documents directly to the platform for further analysis.</p><figure><img alt="LUIS APP flow structure:" src="https://cdn-images-1.medium.com/max/875/1*D_nhWHg2C4sYZ5PtIotkdg.png" /><figcaption>fig:1 MSLUIS Architecture</figcaption></figure><p>To integrate ServiceNow with MSLUIS, you need to complete the following steps:</p><blockquote>1. Sign into your Microsoft Azure portal, navigate to Cognitive Services and create a LUIS application. <br> 2. Once your LUIS application is created, you’ll need to train it by adding intents, utterances, and entities. <br> 3. Once the LUIS app is trained, you’ll need to generate an API key. <br> 4. Create Instance in service now &amp; incident management in service now and generate API’s.<br> 5. Configure the MSLUIS ServiceNow Integration App by entering the API key generated in step 3. <br> 6. write python code for user quarries solved by chatbot for.e.g.fetch &amp; raise the incident etc.<br> 7. deploy chatbot in Azure Service. e.g. Ubuntu server<br> 8. Test the agent by adding some user input to the chatbot and verifying the results.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SNOUPSRu2na5T2F8bEUnrw.png" /><figcaption>fig:2 Solution Architecture of Start to End process of MSLUIS API Integration with Service Now API.</figcaption></figure><p><strong>Results:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/585/1*TsOwqlZv6c93oAEx9f0FLA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/296/1*2cSO1DZgkt1-KtKQCQnEKw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZZvqqBmXEaUmgNOevLk2Yg.png" /><figcaption>service now virtual agent</figcaption></figure><p>Next steps: Sentiment Analysis based chatbot capability to understand sentiment analysis. It can detect the sentiment of a given text and classify it into Positive, Negative and Neutral categories of user queries.</p><h3>Thanks for reading!…</h3><p>If you thought this was interesting, leave a clap or two, and subscribe for future updates!!!</p><p>You can subscribe to my<a href="https://medium.com/datadreamers/newsletters/data-dreamers"> <strong>Medium Newsletter</strong></a><strong> to stay tuned and receive my content</strong>. <em>I promise it will be unique!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1fe82b51e035" width="1" height="1" alt=""><hr><p><a href="https://medium.com/datadreamers/intelligent-luis-based-chatbot-integration-with-service-now-1fe82b51e035">Revolutionizing Customer Service: How Intelligent Luis Chatbot Integration with ServiceNow is…</a> was originally published in <a href="https://medium.com/datadreamers">DataDreamers</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>