<?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 Geetha on Medium]]></title>
        <description><![CDATA[Stories by Geetha on Medium]]></description>
        <link>https://medium.com/@bgeethasagar?source=rss-2355e2facfef------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Stories by Geetha on Medium</title>
            <link>https://medium.com/@bgeethasagar?source=rss-2355e2facfef------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 02:44:41 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bgeethasagar/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[How to Connect PostgreSQL (via Docker) to Tableau on a Mac — Live & Local]]></title>
            <link>https://medium.com/@bgeethasagar/how-to-connect-postgresql-via-docker-to-tableau-on-a-mac-live-local-e47290386214?source=rss-2355e2facfef------2</link>
            <guid isPermaLink="false">https://medium.com/p/e47290386214</guid>
            <category><![CDATA[postgresql]]></category>
            <category><![CDATA[postgres]]></category>
            <category><![CDATA[docker]]></category>
            <category><![CDATA[tableau]]></category>
            <category><![CDATA[data-visualization]]></category>
            <dc:creator><![CDATA[Geetha]]></dc:creator>
            <pubDate>Sat, 12 Jul 2025 19:19:56 GMT</pubDate>
            <atom:updated>2025-07-12T19:19:56.537Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to Connect PostgreSQL (via Docker) to Tableau on a Mac — Live &amp; Local</h3><p>While working on a data analytics project involving air quality data, I needed to visualize insights from a <strong>PostgreSQL database running locally in Docker</strong> — and connect it to <strong>Tableau Desktop on a Mac</strong> using a <strong>live connection</strong>. There were a few tricky parts, so here’s a step-by-step guide that might save you time.</p><h3>My Setup</h3><ul><li><strong>macOS (M1/M2 or Intel)</strong></li><li><strong>PostgreSQL 14 in Docker</strong></li><li><strong>Tableau Desktop for Mac</strong></li><li><strong>pgAdmin (optional for browsing DB)</strong></li><li><strong>Docker Compose + .env file</strong></li></ul><h3>Step 1: Run PostgreSQL in Docker</h3><p>Here’s a sample docker-compose.yml to spin up PostgreSQL locally:</p><pre>services:<br>  postgres:<br>    image: postgres:14<br>    container_name: my-postgres-container<br>    ports:<br>      - &quot;5433:5432&quot;  # Host maps 5433 to container&#39;s 5432<br>    environment:<br>      POSTGRES_DB: &lt;your_database_name&gt;<br>      POSTGRES_USER: &lt;your_username&gt;<br>      POSTGRES_PASSWORD: &lt;your_password&gt;<br>    volumes:<br>      - postgres-data:/var/lib/postgresql/data<br>volumes:<br>  postgres-data:</pre><p>In your .env file or environment variables, define:</p><pre>POSTGRES_DB=your_database_name<br>POSTGRES_USER=your_username<br>POSTGRES_PASSWORD=your_password</pre><p>Once the container is running, you can access PostgreSQL via tools like psql, pgAdmin, or connect directly from Tableau.</p><h3>Step 2: Create Your Database (if not already created)</h3><p>If your database wasn’t initialized by Docker, connect using psql or pgAdmin and run:</p><pre>CREATE DATABASE market_research;</pre><p>Or name it anything relevant to your project.</p><h3>Step 3: Download PostgreSQL JDBC Driver</h3><p>To allow Tableau to talk to PostgreSQL, download the JDBC driver:</p><ol><li>Visit: <a href="https://jdbc.postgresql.org/download/">https://jdbc.postgresql.org/download/</a></li><li>Download the <strong>JAR file</strong> (e.g., version 42.7.7 for Java 8+)</li><li>Move the JAR to this location on your Mac:</li></ol><pre>~/Library/Tableau/Drivers</pre><p>(You may need to create this folder if it doesn’t exist.)</p><h3>Step 4: Connect Tableau to PostgreSQL</h3><p>In <strong>Tableau Desktop</strong>, follow these steps:</p><ol><li>Choose <strong>PostgreSQL</strong> as your data source.</li><li>Enter connection details:</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dxoQcUOupJJaFzkdDaejrQ.png" /><figcaption>db conection parameters</figcaption></figure><h3>Troubleshooting Tips</h3><ul><li>Run docker ps to confirm your container is up and PostgreSQL is running.</li><li>Use psql or pgAdmin to verify that your database exists.</li><li>Make sure the port (5433) isn’t blocked by your firewall or used by another app.</li><li>Avoid using pgBouncer or pooled connections — Tableau needs <strong>stateful connections</strong>.</li></ul><h3>Why I Did This</h3><p>My project involved heavy SQL work inside PostgreSQL and dashboard storytelling in Tableau. Instead of exporting CSVs every time, a <strong>live connection</strong> saved me hours and let me iterate faster — especially useful for:</p><ul><li>KPI dashboards</li><li>Geo-level visualizations (e.g., AQI by city)</li><li>Trend charts that refresh with each query</li></ul><h3>Final Thoughts</h3><p>If you’re building SQL projects locally with Docker and want to visualize them in Tableau, this setup is efficient, lightweight, and easy to replicate.</p><p><strong>Happy SQLing!</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e47290386214" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Can SLMs challenge LLM Giants like ChatGPT-4 & DeepSeek?]]></title>
            <link>https://medium.com/@bgeethasagar/can-slms-challenge-llm-giants-like-chatgpt-4-deepseek-f82c9f515d95?source=rss-2355e2facfef------2</link>
            <guid isPermaLink="false">https://medium.com/p/f82c9f515d95</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[deepseek]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[small-language-model]]></category>
            <category><![CDATA[large-language-models]]></category>
            <dc:creator><![CDATA[Geetha]]></dc:creator>
            <pubDate>Sun, 15 Jun 2025 06:08:45 GMT</pubDate>
            <atom:updated>2025-06-15T06:08:45.506Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s181lnUhyBI8IHNF_F524g.png" /></figure><h3>Introduction</h3><p>The AI industry has been obsessed with bigger, more powerful models. Large Language Models (LLMs) like ChatGPT, DeepSeek, and Claude have dominated discussions, pushing the limits of AI capabilities. However, a new trend is gaining traction—Small Language Models (SLMs), which focus on efficiency, cost reduction, and accessibility.</p><p>In this blog, we explore the rise of SLMs, their advantages over LLMs, and how the AI market is reacting to this trend.</p><p><strong>Are Small Language Models the Future? How SLMs Are Challenging LLM Giants ?</strong></p><h3>LLMs vs. SLMs: The Efficiency War</h3><h3>Large Language Models (LLMs): The Powerhouses</h3><p>LLMs are built on <strong>massive datasets</strong> with <strong>billions of parameters</strong>, allowing them to generate coherent, high-quality responses. However, their growth has led to challenges:</p><p><strong>Pros:</strong></p><ul><li>Superior reasoning &amp; problem-solving.</li><li>Can process vast amounts of data.</li><li>General-purpose capabilities.</li></ul><p><strong>Cons:</strong></p><ul><li><strong>Expensive</strong>: Training and inference require high-end GPUs.</li><li><strong>Slow latency</strong>: Not suitable for real-time applications.</li><li><strong>High energy consumption</strong>: Not sustainable for long-term scaling.</li></ul><p><strong>Examples:</strong> OpenAI’s GPT-4, Google Gemini, and DeepSeek’s large-scale models.</p><h3>Small Language Models (SLMs): The Underdogs</h3><p>SLMs take a different approach—trading some power for <strong>speed, efficiency, and affordability</strong>. Instead of relying on massive compute power, they optimize performance with fewer parameters and <strong>domain-specific tuning</strong>.</p><p><strong>Pros:</strong></p><ul><li>Faster inference times.</li><li>Lower compute requirements (runs on CPUs and edge devices).</li><li>More cost-effective for businesses and startups.</li></ul><p><strong>Cons:</strong></p><ul><li>Limited knowledge base compared to LLMs.</li><li>Struggles with long-form reasoning.</li><li>Needs fine-tuning for high accuracy in complex tasks.</li></ul><p><strong>Examples:</strong> Mistral 7B, Phi-2 (Microsoft), TinyLlama, DistilBERT.</p><h3>Market Trends: DeepSeek, ChatGPT, and NVIDIA’s Challenges</h3><h3>DeepSeek’s Open-Source Disruption</h3><p>DeepSeek, a <strong>Chinese AI company</strong>, is making waves with open-source LLMs, directly competing with OpenAI and Google. The <strong>key differentiator</strong>? <strong>Smaller, fine-tuned models</strong> that balance cost and performance.</p><p><strong>Why It Matters:</strong></p><ul><li>Open-source <strong>lowers entry barriers</strong> for AI development.</li><li><strong>Businesses prefer tailored, domain-specific models</strong> over massive general-purpose ones.</li><li><strong>China’s focus on AI independence</strong> is driving efficiency-oriented innovation.</li></ul><h3>NVIDIA’s Market Challenges: Is the GPU Craze Slowing Down?</h3><p>For years, NVIDIA has been the backbone of AI, providing <strong>high-performance GPUs</strong> crucial for LLM training. However, recent trends suggest a shift:</p><p><strong>Stock fluctuations:</strong> As SLMs reduce dependency on large-scale compute, fewer GPUs are required.</p><p><strong>The Rise of AI Chips:</strong> Companies like <strong>Apple (M-series), Google (TPUs), and Tesla (Dojo)</strong> are investing in custom AI chips that challenge NVIDIA’s market.</p><p><strong>The Energy Debate:</strong> LLMs require enormous electricity, pushing AI companies toward <strong>low-power alternatives</strong> like SLMs.</p><h3>The Future: Will SLMs Replace LLMs?</h3><p>While <strong>SLMs are not replacing LLMs entirely</strong>, they are carving out their own niche. The future likely involves a <strong>hybrid AI approach</strong>:</p><ul><li><strong>SLMs for real-time, on-device applications</strong> (smartphones, IoT, and enterprise AI tools).</li><li><strong>LLMs for complex, large-scale reasoning tasks</strong> (AGI, deep research, content generation).</li><li><strong>Efficient AI compute</strong> will become a priority, benefiting <strong>SLMs over time</strong>.</li></ul><p><strong>What’s Next?</strong></p><ul><li>AI companies may prioritize <strong>fine-tuned SLMs over massive LLMs</strong>.</li><li>More businesses will adopt <strong>customized models over general-purpose AI</strong>.</li><li>NVIDIA must <strong>adapt to the efficiency-first AI era</strong> or risk losing market share.</li></ul><h3>Conclusion</h3><p>As AI progresses, <strong>bigger isn’t always better</strong>. SLMs offer a <strong>scalable, cost-effective</strong>, and <strong>energy-efficient</strong> alternative to LLMs, making them a strong contender in the evolving AI ecosystem. Whether NVIDIA, OpenAI, and Google fully embrace this trend or resist it remains to be seen. But one thing is clear—<strong>the AI landscape is shifting.</strong></p><p><strong>What are your thoughts?</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f82c9f515d95" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Hi, I’m Geetha — And This Is How Data Became My Language…]]></title>
            <link>https://medium.com/@bgeethasagar/hi-im-geetha-and-this-is-how-data-became-my-language-f0bda52f8050?source=rss-2355e2facfef------2</link>
            <guid isPermaLink="false">https://medium.com/p/f0bda52f8050</guid>
            <category><![CDATA[introduction]]></category>
            <category><![CDATA[first-post]]></category>
            <category><![CDATA[about-me]]></category>
            <dc:creator><![CDATA[Geetha]]></dc:creator>
            <pubDate>Sun, 15 Jun 2025 05:59:36 GMT</pubDate>
            <atom:updated>2025-10-03T02:16:59.664Z</atom:updated>
            <content:encoded><![CDATA[<h3>Hi, I’m Geetha — And This Is How Data Became My Language…</h3><p>I’ve always been someone who loves figuring out how things work. Whether it was building random contraptions as a kid or dissecting the logic behind a football game, curiosity has always been my starting point. Over time, that curiosity led me to the world of data—not just numbers, but patterns, behaviors, decisions, and stories waiting to be told.</p><p>Today, I work at the intersection of data, business, and product. I’ve spent over three years diving into supply chain problems, marketing insights, and nonprofit outreach—using tools like Python, SQL, and Tableau to help teams make smarter decisions. From reducing churn to improving inventory flow, I’ve come to appreciate the clarity that comes from combining structured thinking with a deep understanding of business needs.</p><p>Currently, I’m pursuing my Master’s in Business Analytics at the University of Cincinnati. This journey has been more than academic—it’s a space where I’m constantly learning how to turn messy datasets into meaningful strategies.</p><p>But there’s more to me than dashboards and models.</p><p>When I log off, you’ll find me on my bike chasing open roads, kicking around a football with friends, or finding my next favorite café (I take coffee seriously). I’m a dog lover—shoutout to Maxie—and I love trying new sports just for the fun of learning something new.</p><p>This blog will be a mix of everything I enjoy: reflections from projects, tips for working with data, and maybe even some insights on how football strategy isn’t all that different from product analytics.</p><p>Thanks for stopping by—and if you’re someone who loves solving problems and learning out loud, I’d love to connect.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f0bda52f8050" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>