<?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 Rajib Bhattacharya on Medium]]></title>
        <description><![CDATA[Stories by Rajib Bhattacharya on Medium]]></description>
        <link>https://medium.com/@rajibbhattacharya?source=rss-aaff8b4add92------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*a1oLCnslPjTGLdP1.</url>
            <title>Stories by Rajib Bhattacharya on Medium</title>
            <link>https://medium.com/@rajibbhattacharya?source=rss-aaff8b4add92------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 03:30:50 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@rajibbhattacharya/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[AI Governance: From AI Experimentation to Enterprise Trust]]></title>
            <link>https://medium.com/@rajibbhattacharya/ai-governance-from-ai-experimentation-to-enterprise-trust-94ab6ff61dba?source=rss-aaff8b4add92------2</link>
            <guid isPermaLink="false">https://medium.com/p/94ab6ff61dba</guid>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai-governance]]></category>
            <category><![CDATA[responsible-ai]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <dc:creator><![CDATA[Rajib Bhattacharya]]></dc:creator>
            <pubDate>Sat, 25 Apr 2026 17:19:48 GMT</pubDate>
            <atom:updated>2026-04-25T17:19:48.604Z</atom:updated>
            <content:encoded><![CDATA[<p>AI Governance is no longer a compliance side-topic. It is becoming one of the most important foundations for scaling <strong>Generative AI, Agentic AI, Machine Learning, and enterprise automation</strong> safely.</p><p>Most organizations are past the “Can we build an AI proof of concept?” stage.</p><p>The bigger question now is:</p><p><strong>Can we trust it in production?</strong></p><p>That is where AI Governance becomes critical.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1013/1*UasToRVWh4sWt3w5yVJ5iA.png" /></figure><h3>What is AI Governance?</h3><p><strong>AI Governance is the operating model that ensures AI systems are safe, compliant, explainable, accountable, secure, and aligned to business value.</strong></p><p>It covers the policies, controls, roles, metrics, architecture, and monitoring practices required to manage AI across its lifecycle.</p><p>In simpler terms:</p><p>AI Governance answers questions like:</p><ul><li>Who owns this AI use case?</li><li>What data is it allowed to use?</li><li>Has the model been tested for accuracy, bias, robustness, and drift?</li><li>Can we explain how the output was generated?</li><li>Is there human oversight for critical decisions?</li><li>What happens when the model fails?</li><li>How do we audit the system later?</li></ul><p>Without clear answers, AI remains an experiment.</p><p>With governance, AI becomes an enterprise capability.</p><h3>Why AI Governance Matters</h3><p>The rise of <strong>LLMs, RAG systems, copilots, autonomous agents, and AI-powered workflows</strong> has created a new type of enterprise risk.</p><p>Traditional IT controls are still necessary, but they are not sufficient.</p><p>AI systems introduce additional challenges:</p><ul><li>Hallucination risk</li><li>Bias and fairness concerns</li><li>Sensitive data leakage</li><li>Unclear accountability</li><li>Model drift</li><li>Prompt injection</li><li>Poor explainability</li><li>Inconsistent outputs</li><li>Third-party model dependency</li><li>Regulatory and audit exposure</li></ul><p>This is especially important for industries like <strong>banking, pharma, healthcare, insurance, manufacturing, retail, and public sector</strong>, where decisions must be traceable, defensible, and compliant.</p><p>In short, AI Governance is not about slowing innovation.</p><p>It is about making innovation sustainable.</p><p>Or, as I like to say:</p><p><strong>AI Governance is not the brake. It is the steering system.</strong></p><p>Because “the model sounded confident” is not an approved control framework. 🙂</p><h3>The Core Pillars of AI Governance</h3><p>A strong AI Governance framework needs to cover both technical and organizational dimensions.</p><h3>1. Strategy and Policy</h3><p>Every organization needs a clear AI policy that defines:</p><ul><li>AI principles</li><li>Acceptable use</li><li>Prohibited use</li><li>Risk appetite</li><li>Ownership model</li><li>Approval process</li><li>Human oversight requirements</li></ul><p>This is where business, technology, risk, legal, compliance, and security teams need to align.</p><p>A good AI policy should not be theoretical. It should translate into practical decision rules for teams building AI solutions.</p><p>For example:</p><ul><li>Which use cases require formal approval?</li><li>Which use cases need human-in-the-loop?</li><li>Which data classes are restricted?</li><li>Which outputs need review before customer or regulatory exposure?</li></ul><h3>2. Data Governance</h3><p>AI quality is directly linked to data quality.</p><p>If the data is poor, stale, biased, unclassified, or poorly governed, the AI system will inherit those weaknesses.</p><p>For enterprise AI, data governance must cover:</p><ul><li>Data quality</li><li>Metadata</li><li>Lineage</li><li>Access controls</li><li>Privacy classification</li><li>Data retention</li><li>Consent and usage rights</li><li>Source-of-truth definition</li></ul><p>This is especially important for <strong>RAG and context engineering</strong>, where the model response depends heavily on retrieved enterprise knowledge.</p><p>A weak knowledge base produces weak AI.</p><p>A governed knowledge base creates more reliable AI.</p><h3>3. Model Risk Management</h3><p>AI models need to be validated before production and monitored after deployment.</p><p>Model Risk Management should include:</p><ul><li>Model documentation</li><li>Intended use definition</li><li>Performance testing</li><li>Bias and fairness testing</li><li>Robustness testing</li><li>Explainability assessment</li><li>Drift monitoring</li><li>Failure mode analysis</li><li>Independent validation for high-risk systems</li></ul><p>For Generative AI, this also includes output quality, grounding, toxicity, hallucination rate, citation accuracy, and resistance to adversarial prompts.</p><h3>4. Responsible AI</h3><p>Responsible AI ensures that AI systems are fair, transparent, explainable, privacy-aware, and aligned with human values.</p><p>It includes:</p><ul><li>Fairness</li><li>Explainability</li><li>Accountability</li><li>Transparency</li><li>Human oversight</li><li>Ethical use</li><li>Inclusive design</li></ul><p>This pillar is important because AI systems are not just technical assets. They influence decisions, experiences, opportunities, and trust.</p><h3>5. Security and Compliance</h3><p>AI expands the enterprise attack surface.</p><p>Security and compliance teams need to consider:</p><ul><li>Prompt injection</li><li>Data exfiltration</li><li>Model abuse</li><li>Identity and access control</li><li>Third-party model risk</li><li>API security</li><li>Sensitive data exposure</li><li>Regulatory obligations</li><li>Audit evidence</li></ul><p>For GenAI systems, it is not enough to secure the application. The entire AI workflow must be secured, including prompts, context, retrieval sources, tools, plugins, APIs, logs, and outputs.</p><h3>6. Operations and Observability</h3><p>Once AI moves into production, governance must become operational.</p><p>This is where <strong>LLMOps, MLOps, monitoring, observability, incident management, and continuous evaluation</strong> become essential.</p><p>Teams should monitor:</p><ul><li>Model performance</li><li>Response quality</li><li>Hallucination rate</li><li>Retrieval accuracy</li><li>Bias signals</li><li>Drift</li><li>Latency</li><li>Cost</li><li>User feedback</li><li>Escalations</li><li>Policy violations</li><li>Incident trends</li></ul><p>AI systems should not be “set and forget.”</p><p>They should be continuously measured and improved.</p><h3>The AI Governance Lifecycle</h3><p>A practical AI Governance lifecycle should look something like this:</p><h3>1. Use Case Intake</h3><p>Start with a structured intake process.</p><p>Capture:</p><ul><li>Business objective</li><li>Expected value</li><li>Users impacted</li><li>Data involved</li><li>Decision impact</li><li>Regulatory sensitivity</li><li>Automation level</li><li>Risk category</li></ul><p>Not every AI use case needs the same level of governance.</p><p>A document summarization tool for internal notes is very different from an AI system that recommends patient treatment, credit decisions, or regulatory submissions.</p><h3>2. Risk Tiering</h3><p>Classify AI use cases based on risk.</p><p>Common risk dimensions include:</p><ul><li>Data sensitivity</li><li>Customer impact</li><li>Legal impact</li><li>Financial impact</li><li>Degree of automation</li><li>Human oversight</li><li>Explainability requirement</li><li>Regulatory exposure</li></ul><p>This enables proportional governance.</p><p>Low-risk use cases should move fast.</p><p>High-risk use cases need stronger controls.</p><h3>3. Data and Model Review</h3><p>Review the data, model, architecture, and intended usage.</p><p>This includes:</p><ul><li>Data source approval</li><li>Data quality checks</li><li>Model selection rationale</li><li>Evaluation approach</li><li>Security review</li><li>Privacy review</li><li>Explainability review</li><li>Vendor risk review, if third-party models are used</li></ul><p>For GenAI, this should also include the prompt design, retrieval design, context engineering approach, and guardrail strategy.</p><h3>4. Guardrails and Testing</h3><p>Before deployment, AI systems should be tested against expected and unexpected behavior.</p><p>Important tests include:</p><ul><li>Accuracy testing</li><li>Bias testing</li><li>Hallucination testing</li><li>Adversarial prompt testing</li><li>Toxicity testing</li><li>Privacy leakage testing</li><li>Edge-case testing</li><li>Human review workflow testing</li><li>Fail-safe behavior testing</li></ul><p>The goal is not perfection.</p><p>The goal is controlled, measurable, and explainable risk.</p><h3>5. Deployment Approval</h3><p>Production deployment should require clear approval criteria.</p><p>Approval should confirm:</p><ul><li>The use case has an owner</li><li>The risk tier is documented</li><li>Required controls are implemented</li><li>Evaluation results are acceptable</li><li>Monitoring is configured</li><li>Incident response is defined</li><li>Human oversight is in place where required</li><li>Audit evidence is available</li></ul><p>This creates accountability.</p><h3>6. Monitoring and Audit</h3><p>After deployment, AI systems need continuous monitoring.</p><p>This is where many organizations struggle.</p><p>A model that performed well during testing may degrade over time due to:</p><ul><li>Data drift</li><li>User behavior changes</li><li>Business rule changes</li><li>Model updates</li><li>Retrieval source changes</li><li>Prompt changes</li><li>External environment changes</li></ul><p>Monitoring should be both technical and business-oriented.</p><p>It is not enough to know that the API is up.</p><p>We need to know whether the AI system is still producing useful, safe, and compliant outcomes.</p><h3>7. Continuous Improvement</h3><p>AI Governance should not be a one-time checklist.</p><p>It should create a feedback loop.</p><p>Use production data, user feedback, incident reports, evaluation results, and audit findings to improve:</p><ul><li>Policies</li><li>Models</li><li>Data quality</li><li>Retrieval pipelines</li><li>Guardrails</li><li>Prompts</li><li>Human review workflows</li><li>Metrics</li><li>Training</li><li>Risk controls</li></ul><p>Good governance gets better with usage.</p><h3>Weak Governance vs Strong Governance</h3><p>Weak AI Governance often looks like this:</p><ul><li>Ad hoc pilots</li><li>No clear ownership</li><li>No policy framework</li><li>Unapproved data usage</li><li>No evaluation metrics</li><li>Limited documentation</li><li>No human oversight</li><li>No monitoring after launch</li><li>No incident response plan</li></ul><p>Strong AI Governance looks different:</p><ul><li>Clear AI policy and accountable owners</li><li>Risk-based control model</li><li>Approved data and model inventory</li><li>Defined evaluation framework</li><li>Security, privacy, and compliance controls</li><li>Human-in-the-loop for critical decisions</li><li>Monitoring and red-teaming</li><li>Audit-ready documentation</li><li>Continuous improvement loop</li></ul><p>This is the difference between AI enthusiasm and AI maturity.</p><h3>Practical AI Governance Metrics</h3><p>Governance should not live only in PowerPoint.</p><p>It needs measurable indicators.</p><p>Useful AI Governance metrics include:</p><ul><li>AI use case approval rate</li><li>Policy compliance rate</li><li>Number of model incidents</li><li>Hallucination rate</li><li>Bias or fairness score</li><li>Retrieval precision</li><li>Response quality score</li><li>Human override rate</li><li>Drift detection frequency</li><li>Audit finding closure time</li><li>Data quality score</li><li>Security exception count</li><li>Cost per AI transaction</li><li>Model performance stability</li><li>Time to approve AI use cases</li></ul><p>The right metrics depend on the industry and use case.</p><p>But the principle is simple:</p><p><strong>If we cannot measure it, we cannot govern it properly.</strong></p><h3>A Practical Operating Model for AI Governance</h3><p>An effective AI Governance model should include clear roles.</p><h3>Executive Steering Committee</h3><p>Owns AI strategy, investment priorities, risk appetite, and enterprise-level accountability.</p><h3>AI Governance Council</h3><p>Defines AI policies, standards, risk tiers, approval processes, and cross-functional governance practices.</p><h3>Data Governance Team</h3><p>Ensures trusted data, metadata, lineage, quality, classification, privacy, and access controls.</p><h3>AI Engineering and Platform Teams</h3><p>Build reusable AI platforms, guardrails, monitoring, evaluation pipelines, model registries, and deployment standards.</p><h3>Risk, Legal, Compliance, and Security Teams</h3><p>Review regulatory, contractual, legal, privacy, cybersecurity, and third-party risks.</p><h3>Business Owners</h3><p>Own the business outcome, usage context, process integration, and user adoption.</p><h3>Human Reviewers</h3><p>Provide domain validation, escalation handling, and oversight for critical or sensitive outputs.</p><p>AI Governance works best when it is federated.</p><p>Central teams define standards.</p><p>Business and technology teams execute with accountability.</p><h3>AI Governance for Generative AI and Agentic AI</h3><p>Generative AI and Agentic AI make governance more complex.</p><p>Traditional machine learning models usually produce predictions.</p><p>GenAI systems produce content, reasoning traces, summaries, recommendations, code, and sometimes actions.</p><p>Agentic AI systems can go further by:</p><ul><li>Calling tools</li><li>Querying systems</li><li>Updating records</li><li>Triggering workflows</li><li>Making multi-step decisions</li><li>Coordinating with other agents</li></ul><p>That means governance must cover not only the model, but also the full AI system.</p><p>For Agentic AI, governance should include:</p><ul><li>Tool access controls</li><li>Action approval rules</li><li>Human-in-the-loop checkpoints</li><li>Audit trails for every action</li><li>Context boundary management</li><li>Memory governance</li><li>Task-level permissions</li><li>Rollback and recovery mechanisms</li><li>Simulation before production execution</li><li>Agent behavior monitoring</li></ul><p>In other words:</p><p><strong>The more autonomy we give AI, the stronger the governance must become.</strong></p><h3>The Role of AI Governance in Business Value</h3><p>AI Governance is sometimes misunderstood as bureaucracy.</p><p>That is a mistake.</p><p>Good governance improves business value because it increases:</p><ul><li>Trust</li><li>Adoption</li><li>Reuse</li><li>Reliability</li><li>Regulatory confidence</li><li>Security</li><li>Operational stability</li><li>Executive sponsorship</li></ul><p>Business users are more likely to adopt AI when they know it is reliable.</p><p>Risk teams are more likely to approve AI when controls are visible.</p><p>Executives are more likely to fund AI when outcomes are measurable.</p><p>Customers are more likely to trust AI when accountability is clear.</p><p>Governance is not the opposite of speed.</p><p>Bad governance slows AI because every team reinvents controls.</p><p>Good governance accelerates AI because teams can reuse approved patterns, platforms, and decision frameworks.</p><h3>Final Thought</h3><p>The AI conversation is shifting.</p><p>Earlier, the focus was:</p><p><strong>“Which model should we use?”</strong></p><p>Now the more important question is:</p><p><strong>“How do we govern the AI system end to end?”</strong></p><p>The winners in enterprise AI will not only be the teams with the most advanced models.</p><p>They will be the teams that can combine:</p><ul><li>Strong data foundations</li><li>Responsible AI principles</li><li>Secure architecture</li><li>Continuous evaluation</li><li>Clear accountability</li><li>Business-aligned governance</li></ul><p>AI Governance is not about saying “no” to innovation.</p><p>It is about making AI trustworthy enough for the enterprise to say “yes.”</p><p>And yes, the model may sound confident.</p><p>But confidence is not governance.</p><p>Controls are. 🙂</p><p>#AIGovernance #ResponsibleAI #GenerativeAI #AgenticAI #EnterpriseAI #AIEngineering #AIArchitecture #DataGovernance #ModelRiskManagement #LLMOps #MLOps #AIEthics #AICompliance #DigitalTransformation #AILeadership</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=94ab6ff61dba" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ Automating Power BI Documentation:]]></title>
            <link>https://medium.com/@rajibbhattacharya/automating-power-bi-documentation-b410a2b8845b?source=rss-aaff8b4add92------2</link>
            <guid isPermaLink="false">https://medium.com/p/b410a2b8845b</guid>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[power-bi]]></category>
            <category><![CDATA[data-analytics]]></category>
            <dc:creator><![CDATA[Rajib Bhattacharya]]></dc:creator>
            <pubDate>Sun, 07 Sep 2025 10:09:17 GMT</pubDate>
            <atom:updated>2025-09-07T10:10:40.225Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Automating Power BI Documentation" src="https://cdn-images-1.medium.com/max/1024/1*5oDz_vsIF3PubW8tjZ9ZgA.jpeg" /></figure><p>Over the years, I’ve noticed a consistent challenge for Power BI developers on my teams — while building reports and dashboards, documentation often takes a back seat. Most end up completing just one document at best, leaving critical details scattered or undocumented.</p><p>I’ve always wished to simplify and automate this process for technical users — and over the past weekend, I built the first version of a tool to do just that.</p><p>👉 Here’s my Streamlit app:</p><p>It accepts a PBIX file and quickly generates an Excel-based documentation pack containing:<br>Model schema details<br>Relationships<br>Power Query details<br>M Parameters<br>DAX tables, measures, and more</p><p>This is powered by the pbixray parser under the hood.</p><p>🔗 GitHub repo (please ⭐️ if you find it useful):<br><a href="https://github.com/rajib2018/pbix-analyzer/">https://github.com/rajib2018/pbix-analyzer/</a></p><p>🔗 Try it out here:<br><a href="https://pbix-analyzer-adv.streamlit.app/">https://pbix-analyzer-adv.streamlit.app/</a></p><p>This is just the start — planning to extend it with more advanced features. Would love to hear your thoughts and suggestions!</p><p><a href="https://www.linkedin.com/search/results/all/?keywords=%23powerbi&amp;origin=HASH_TAG_FROM_FEED">#PowerBI</a> <a href="https://www.linkedin.com/search/results/all/?keywords=%23dataanalytics&amp;origin=HASH_TAG_FROM_FEED">#DataAnalytics</a> <a href="https://www.linkedin.com/search/results/all/?keywords=%23streamlit&amp;origin=HASH_TAG_FROM_FEED">#Streamlit</a> <a href="https://www.linkedin.com/search/results/all/?keywords=%23pythonfordata&amp;origin=HASH_TAG_FROM_FEED">#PythonForData</a> <a href="https://www.linkedin.com/search/results/all/?keywords=%23automation&amp;origin=HASH_TAG_FROM_FEED">#Automation</a> <a href="https://www.linkedin.com/search/results/all/?keywords=%23dataengineering&amp;origin=HASH_TAG_FROM_FEED">#DataEngineering</a> <a href="https://www.linkedin.com/search/results/all/?keywords=%23opensource&amp;origin=HASH_TAG_FROM_FEED">#OpenSource</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b410a2b8845b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[R Data Viz — Exploring ggplot2 — Part 1]]></title>
            <link>https://medium.com/@rajibbhattacharya/r-data-viz-exploring-ggplot2-part-1-b183d4768571?source=rss-aaff8b4add92------2</link>
            <guid isPermaLink="false">https://medium.com/p/b183d4768571</guid>
            <category><![CDATA[r-programming]]></category>
            <category><![CDATA[dataviz]]></category>
            <category><![CDATA[ggplot2]]></category>
            <category><![CDATA[data-analytics]]></category>
            <category><![CDATA[data-visualization]]></category>
            <dc:creator><![CDATA[Rajib Bhattacharya]]></dc:creator>
            <pubDate>Tue, 31 May 2022 07:35:51 GMT</pubDate>
            <atom:updated>2022-05-31T07:39:08.933Z</atom:updated>
            <content:encoded><![CDATA[<p>R Data Viz — Exploring ggplot2 — Part 1</p><p><strong><em>Moving my old blog post to Medium</em></strong></p><p>Data visualizations are for business users to see the visual representation of Analytics including charts, graphs, maps and other graphical formats. Data visualizations illustrate difficult concepts, unearth relationships among different data elements and also help in spotting hidden trends and patterns within the data set.</p><p>R is a very powerful open source programming language and software environment for statistical computing and graphics. It compiles and runs on various UNIX platforms, Windows and MacOS.</p><p>The grammar of graphics based plotting system (ggplot2) is the starting point for data visualizations or graphics in R. You can use ggplot2 to plot complex multi-layered graphs with ease.</p><p>I will be using the following R version in my examples:</p><p><strong>R version 3.4.0 (2017–04–21) — “You Stupid Darkness”</strong></p><p>Here are some examples to get you started –</p><p>If you do not have ggplot2 installed, you can use the following syntax for install ggplot2:</p><p><strong>install.packages(“ggplot2”)</strong></p><p>To load the ggplot2 libraries, use the following syntax:</p><p><strong>library(ggplot2)</strong></p><p>I will be using the famous mtcars (or Motor Trend Car Road Tests) dataset which is one of preloaded datasets in R. To find more details about the <strong>mtcars</strong> dataset, please check the following link:</p><p><a href="https://stat.ethz.ch/R-manual/R-devel/library/datasets/html/mtcars.html">R: Motor Trend Car Road Tests</a></p><p>You can use the following syntax for checking mtcars dataset:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/583/0*i0bHSOhjK41nbN7I" /></figure><p>Let us first try to plot the columns miles per gallon (mpg) and displacement (disp) using the <strong>qplot</strong> (quick plot) function:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/709/0*_PcbhO5kyXKDmlmP" /></figure><p>Here is another variant of the qplot function for the above plot:</p><p><strong>qplot(mpg, disp, data = mtcars)</strong></p><p>You can realize the difference between qplot() and the basic plot() function in R if you see the output for these two columns using plot() function. The syntax to be used is:</p><p><strong>plot(mtcars$mpg, mtcars$disp)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/689/0*bWtnzhOdL-rDyy94" /></figure><p>To understand the power of power of ggplot2, let us try to customize the visualization using <strong>qplot</strong>. Let us use the following syntax to add colors to a basic scatter plot:</p><p><strong>qplot(mpg, disp, data = mtcars, color = cyl)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/702/0*RnWp1kkL9wYX-NpF" /></figure><p>Here we are using the argument color with the qplot function and we are using the no. of cylinders (cyl) column for changing these colors. As you can see the colors used in the scatter plot are different shades of Blue. If you need to use different colors for different groups, you can use the following syntax:</p><p><strong>qplot(mpg, disp, data = mtcars, colour=factor(cyl))</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/703/0*9_R4B-1iYdE-CdL9" /></figure><p>Now to change the sizes of points for the cyl column, you can also use the following syntax:</p><p><strong>qplot(mpg, disp, data = mtcars, colour=factor(cyl), size=cyl)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/703/0*mu-Wqs68-X5HHKXo" /></figure><p>Also, to add different shapes for different values of cyl column the following syntax can be tried out:</p><p><strong>qplot(mpg, disp, data = mtcars, colour=factor(cyl), size=cyl, shape=factor(cyl))</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/706/0*kkzgfvWn49DJlJtg" /></figure><p>To provide a customized title for your chart, you can use the <strong>main</strong> argument of the qplot function:</p><p><strong>qplot(mpg, disp, data = mtcars, colour=factor(cyl), size=cyl, shape=factor(cyl), main=”MTCars”)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*-b1cPiZpYeZJkq2Y" /></figure><p>That covers the basics of ggplot2. I am going to cover more features and different types of charts in the next blog post. Stay tuned ….</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b183d4768571" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Data Projects to Data Products]]></title>
            <link>https://medium.com/@rajibbhattacharya/data-projects-to-data-products-974360c1e83b?source=rss-aaff8b4add92------2</link>
            <guid isPermaLink="false">https://medium.com/p/974360c1e83b</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[product-data]]></category>
            <category><![CDATA[data-product-creation]]></category>
            <category><![CDATA[product-thinking]]></category>
            <category><![CDATA[data-product-manager]]></category>
            <dc:creator><![CDATA[Rajib Bhattacharya]]></dc:creator>
            <pubDate>Mon, 23 May 2022 07:27:49 GMT</pubDate>
            <atom:updated>2022-05-23T17:03:42.342Z</atom:updated>
            <content:encoded><![CDATA[<p>Converting Projects to Products is a relatively new concept. With the current situation in the technology industry all organizations are looking to convert their projects to assets or products so that these assets / products can we reused within the organization instead of creating another solution silo. <strong><em>Product Thinking</em></strong> has become the new normal.</p><p>When it comes to Data, it is very important that the Data Assets / Products are created and maintained within the organization to break the barriers between different departments and creating a generic dataset / data asset / data product for the entire organization. Organizations are measuring and trying to measure / understand the ROI of the data products that they create. As we are seeing more and more Data &amp; Analytics use cases, most of the organizations are looking to create data-based / insight-based business models which are future-oriented. However, this is not an easy task.</p><p>The Methodology has three parts —</p><p><strong>Part 1: </strong>If you solve a business problem with data, it should be clearly documented — the code, algorithms &amp; data elements used. The lineage of data and the metadata are of paramount importance. If the data products capture important metrics / KPIs, the metrics should be a part of organization’s <strong><em>Metrics Repository</em></strong>. This repository will include the metadata of metrics / KPIs (business definitions, technical definitions, data elements etc.) being used in all data products.</p><p><strong>Part 2:</strong> Data product managers / Data leaders need to start looking for options to make the data product / asset generic, so that it can used for a wider audience (preferably the entire organization).</p><p><strong>Part 3:</strong> Once significant investments have been made on these data products, the organization will be looking to reuse these data assets / data products for related business problems from different departments which will ultimately reduce the need of data investments for projects.</p><p>#DataProducts #DataProductCreation #DataProductManagement #ProductThinking #ProductManagement</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=974360c1e83b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>