<?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 Upkar Lidder on Medium]]></title>
        <description><![CDATA[Stories by Upkar Lidder on Medium]]></description>
        <link>https://medium.com/@lidderupk?source=rss-50c02028dfb6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*zIBQheq7mlysZeKO_zUCfQ.png</url>
            <title>Stories by Upkar Lidder on Medium</title>
            <link>https://medium.com/@lidderupk?source=rss-50c02028dfb6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 20:39:03 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@lidderupk/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[If serverless apps are stateless, how do you store data when building functions?]]></title>
            <link>https://medium.com/@lidderupk/if-serverless-apps-are-stateless-how-do-you-store-data-when-building-functions-9cab25394a97?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/9cab25394a97</guid>
            <category><![CDATA[openwhisk]]></category>
            <category><![CDATA[apache-openwhisk]]></category>
            <category><![CDATA[serverless]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Sat, 07 Dec 2019 00:42:13 GMT</pubDate>
            <atom:updated>2019-12-07T00:42:13.115Z</atom:updated>
            <content:encoded><![CDATA[<p>Question asked on slack — <a href="https://openwhisk.apache.org/"><strong><em>OpenWhisk</em></strong></a><strong><em> documentation states that serverless architecture fits microservices. But microservices have to store data somewhere. Because container can be stopped at any time I can’t use connection pooling or something similar.</em></strong></p><p>So this got “me” thinking …</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*33AXSY2Q6BjJ1mgMtnF5ew.png" /><figcaption>Obviously thinking about how to store state in serverless. Also, not me!</figcaption></figure><p>Yes, the idea is for the functions to be stateless. This does not mean that they cannot read / write data. You could have a function/microservice that reads data from a database and another function/microservice that writes data to a database (like nosql or rdbms). You can also read data from write data to a file storage system (like s3 or IBM COS). <a href="https://medium.com/u/604ac4ebfc5f">James Thomas</a> wrote a <a href="http://jamesthom.as/blog/2018/05/31/using-cloud-object-storage-from-ibm-cloud-functions-node-dot-js/">wonderful article on this topic</a>.</p><p>Another architecture would be to use some of the pre-built packages in openWhisk like <a href="https://github.com/apache/openwhisk-package-cloudant">openwhisk-package-cloudant</a> to talk to some persistence solution.</p><p>Lastly, you can use a message hub like <a href="https://github.com/apache/openwhisk-package-kafka">Kafka</a> to loosely couple the different microservices so that a function could process data and another would interface with the storage solution once the processing is done based on messages arriving on some queue. This ensures the different components are very loosely coupled.</p><p>When you say container can be stopped at any time, there is an upper limit in OpenWhisk of how long a function can run. This is listed in the docs <a href="https://github.com/apache/openwhisk/blob/master/docs/reference.md#system-limits">as 5 mins</a>. The hosted version on <a href="https://cloud.ibm.com/docs/openwhisk?topic=cloud-functions-limits">IBM Cloud increases this to 10 mins</a>. Of course a function may stop if it fails due to a runtime error before writing data out.</p><p>Here are some code patterns that use functions with databases and cloud object store.</p><ul><li><a href="https://developer.ibm.com/patterns/serverless-event-stream-processing/">Deploy serverless functions that respond to messages and handle streams</a></li><li><a href="https://developer.ibm.com/patterns/serverless-cloud-object-storage-image-processing/">Serverless image processing with Cloud Object Storage</a></li><li><a href="https://developer.ibm.com/patterns/serverless-microservices-api-gateway/">Use a serverless, event-driven architecture to run code that scales automatically</a></li></ul><p><a href="https://medium.com/u/bf01a11701fe">Max Katz</a> — thoughts?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9cab25394a97" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[It’s now even easier to get started with Serverless using the new standalone Apache OpenWhisk!]]></title>
            <link>https://medium.com/@lidderupk/its-now-even-easier-to-get-started-with-serverless-using-the-new-standalone-apache-openwhisk-acef95aa979c?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/acef95aa979c</guid>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[openwhisk]]></category>
            <category><![CDATA[apache-openwhisk]]></category>
            <category><![CDATA[ibm]]></category>
            <category><![CDATA[apache]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Sat, 16 Nov 2019 00:01:23 GMT</pubDate>
            <atom:updated>2019-11-16T00:01:23.520Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dgYWPCF8TBKG9om93tF5Lg.png" /><figcaption>Subset of OpenWhisk runtimes</figcaption></figure><p>One of the cool things about OpenWhisk was the ability to start a local copy and get coding very quickly. <a href="https://medium.com/u/604ac4ebfc5f">James Thomas</a> has an awesome post on this — <a href="http://jamesthom.as/blog/2018/01/19/starting-openwhisk-in-sixty-seconds/">Starting OpenWhisk in Sixty Seconds</a>. This also enabled developers to create and test their Serverless solutions on their machines. The amazing OpenWhisk community have taken this one step further. <a href="https://twitter.com/chetanmeh">Chetan Mehrotra</a> recently added code to enable running OpenWhisk as a standalone jar! How cool is that!</p><p>At a high level, they have taken out CouchDB and Kafka and replaced them with in memory persistence layer and a queueing system. The controller and invoker have also been slimmed down. There is obviously more to it and is well captured in the following links.</p><ul><li><a href="https://github.com/apache/openwhisk/pull/4516#4216">https://github.com/apache/openwhisk/pull/4516</a></li><li><a href="https://github.com/apache/openwhisk/pull/4516#4216">https://github.com/apache/openwhisk/pull/4216</a></li><li><a href="https://lists.apache.org/thread.html/7425131f1fc11a9fd21e9c049be702837841c47004da03b7f215a0d6@%3Cdev.openwhisk.apache.org%3E">https://lists.apache.org/thread.html/7425131f1fc11a9fd21e9c049be702837841c47004da03b7f215a0d6@%3Cdev.openwhisk.apache.org%3E</a></li></ul><p>So how do you get started?</p><h4>Step 1: Build the jar file</h4><ol><li>You can build it yourself by following the steps in the <a href="https://github.com/apache/openwhisk/tree/master/core/standalone">official repo</a>. The final jar will be available in the /bin folder. This is the preferable approach as you always get the latest features</li><li>Alternatively, if you must insist, you can download the <a href="https://github.com/chetanmeh/incubator-openwhisk/releases/tag/0.14">pre-built jar file from here</a>. I am not sure if this will be kept up to date. So try at your own risk!</li></ol><h4>Step 2: run the jar file</h4><p>Once you have the jar file, you can run it as follows</p><p>java -jar bin/openwhisk-standalone.jar</p><p>There are numerous other options available, but this will suffice for now. That’s it! You have a Serverless platform running on your local machine. You can run all your favorite wsk commands just as your would with a production level distributed OpenWhisk installation!</p><h4>Step 3: create and deploy a simple action</h4><p>If you have never deployed an action on Apache OpenWhisk or IBM Cloud, you can follow these steps</p><p>3.1 Download the <a href="https://github.com/apache/openwhisk-cli#where-to-download-the-binary-of-openwhisk-cli">wsk cli</a>.</p><p>3.2 Create your function</p><pre>function main(args) {<br>  if (args &amp;&amp; args.name) {<br>    console.log(`hello ${args.name}`);<br>    return { msg: `hello ${args.name}` };<br>  } else {<br>    console.log(`hello world`);<br>    return { msg: `hello world` };<br>  }<br>}</pre><p>3.3 Before you deploy your function/action, you need to set the auth property using the wsk cli. This command was provided to you when you started the jar file. Simply copy paste in terminal!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8S4Pxu1F8r0uZo15IYdSZg.png" /></figure><p>4. That’s it! Let’s deploy the index.js file as an action</p><pre>$ wsk action create hello index.js<br>ok: created action hello</pre><p>We can now invoke this action</p><pre>$ wsk action invoke hello -r<br>{<br>    &quot;msg&quot;: &quot;hello world&quot;<br>}</pre><pre># with params<br>$ wsk action invoke hello -r -p name upkar<br>{<br>    &quot;msg&quot;: &quot;hello upkar&quot;<br>}</pre><p>If you were paying close attention you would have noticed that when you run the jar file, it opens a browser with the new OpenWhisk Function Playground! This makes it even easier to write and test your functions if you are new to OpenWhisk. Go ahead and try it out!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gBsfxZvJqNOPeh9B9ZNleg.png" /><figcaption>OpenWhisk Function Playground</figcaption></figure><p>How cool was that! If you like this, <a href="https://github.com/apache/openwhisk">please give the repo some love</a>! More to come on how this blazingly fast jar can help developers in their Serverless journeys.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=acef95aa979c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Gentle Introduction to Apache Spark for Developers]]></title>
            <link>https://medium.com/@lidderupk/a-gentle-introduction-to-apache-spark-for-developers-d169aee57c3a?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/d169aee57c3a</guid>
            <category><![CDATA[spark]]></category>
            <category><![CDATA[apache]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Wed, 21 Aug 2019 21:53:26 GMT</pubDate>
            <atom:updated>2019-08-21T21:53:26.193Z</atom:updated>
            <content:encoded><![CDATA[<p>I recently hosted an <a href="https://www.crowdcast.io/e/8u3d6q3c">online meetup on Apache Spark</a> with <a href="https://medium.com/u/262975298e3a">IBM Developer</a>. Spark has been around for a few years, but the <strong><em>interest in still growing to my surprise</em></strong>. Apache Spark was developed at the University of California, Berkeley’s AMPLab. The Spark codebase was open sourced and donated to the Apache Software Foundation in 2010.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*agv4xx-9qWAWzLEFfiXj9g.png" /><figcaption>Introduction to Apache Spark and PixieDust</figcaption></figure><p>The background of the attendees was quite diverse</p><ul><li><strong>Developer (25%)</strong></li><li><strong>Architect (12.5%)</strong></li><li><strong>Data Scientist (41.7%)</strong></li><li><strong>Other (12.8%)</strong></li></ul><p>We looked at the <strong>WHAT</strong> and <strong>WHY</strong> of spark and then dove in the three data structures that you might encounter when working with Spark …</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/970/1*KQYeTwCltFJBca0UZtD9-A.png" /></figure><p>We also looked at some Transformations, Actions, built-in functions and UDF (user defined functions). For example the following function creates a new column called <strong>GENDER </strong>based on the contents of the column <strong>GenderCode.</strong></p><pre><em># ------------------------------</em><br><em># Derive gender from salutation</em><br><em># ------------------------------</em><br><strong>def</strong> deriveGender(col):<br>    <em>&quot;&quot;&quot; input: pyspark.sql.types.Column</em><br><em>        output: &quot;male&quot;, &quot;female&quot; or &quot;unknown&quot;</em><br><em>    &quot;&quot;&quot;</em>    <br>    <strong>if</strong> col <strong>in</strong> [&#39;Mr.&#39;, &#39;Master.&#39;]:<br>        <strong>return</strong> &#39;male&#39;<br>    <strong>elif</strong> col <strong>in</strong> [&#39;Mrs.&#39;, &#39;Miss.&#39;]:<br>        <strong>return</strong> &#39;female&#39;<br>    <strong>else</strong>:<br>        <strong>return</strong> &#39;unknown&#39;;<br>    <br>deriveGenderUDF = func.udf(<strong>lambda</strong> c: deriveGender(c), types.StringType())<br>customer_df = customer_df.withColumn(&quot;GENDER&quot;, deriveGenderUDF(customer_df[&quot;GenderCode&quot;]))<br>customer_df.cache()</pre><p><strong>withColumn</strong> creates a new column in the customer_df dataframe with the values from <strong>deriveGenderUDF</strong> (our user defined function). The deriveGenderUDF is essentially the <strong>deriveGender</strong> function. If this does not make sense, watch the webinar as we go into a lot more detail.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/981/1*FTBmSEX9V89ZrG2ya77uYA.png" /></figure><p>Finally, we created a spark cluster environment on IBM Cloud, and used a Jupyter notebook to explore customer data with the following columns …</p><pre>&quot;CUST_ID&quot;, <br>&quot;CUSTNAME&quot;, <br>&quot;ADDRESS1&quot;, <br>&quot;ADDRESS2&quot;, <br>&quot;CITY&quot;, <br>&quot;POSTAL_CODE&quot;, <br>&quot;POSTAL_CODE_PLUS4&quot;, <br>&quot;STATE&quot;, <br>&quot;COUNTRY_CODE&quot;, <br>&quot;EMAIL_ADDRESS&quot;, <br>&quot;PHONE_NUMBER&quot;,<br>&quot;AGE&quot;,<br>&quot;GenderCode&quot;,<br>&quot;GENERATION&quot;,<br>&quot;NATIONALITY&quot;, <br>&quot;NATIONAL_ID&quot;, <br>&quot;DRIVER_LICENSE&quot;</pre><p>After cleaning the data using in-built and user defined methods, we used PixieDust to visualize the data. The cool thing about pixieDust is that you don’t need to set it up or configure it. You just pass it a Spark DataFrame or a Pandas DataFrame and you are good to go! You can find the complete <a href="https://github.com/IBM/analyze-customer-data-spark-pixiedust">notebook here</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*VbfDDknhZAA_mqzSzTb0zw.png" /></figure><p>Thank you <a href="https://medium.com/u/262975298e3a">IBM Developer</a> and <a href="https://medium.com/u/bf01a11701fe">Max Katz</a> for the opportunity to present and special thanks to Lisa Jung for being a patient co-presenter!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tP-dw4Oj_42BYbkdtYbjMA.png" /><figcaption>Apache Spark</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d169aee57c3a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Polyglot with Open Source and Serverless]]></title>
            <link>https://medium.com/@lidderupk/polyglot-with-open-source-and-serverless-aa38a7844926?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa38a7844926</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[openwhisk]]></category>
            <category><![CDATA[ibm]]></category>
            <category><![CDATA[serverless]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Sat, 03 Aug 2019 18:38:19 GMT</pubDate>
            <atom:updated>2019-08-03T18:38:19.234Z</atom:updated>
            <content:encoded><![CDATA[<p>I often mentor at hackathons as part of my job as a developer advocate at IBM and one of the common problems teams face is which language to pick for their solution. Generally the person who originally thought of the idea will decide on a language. That is better than analysis paralysis, however there is a better way! If you can divide your application into multiple services and/or functions, you can use a polyglot framework such as openwhisk.</p><blockquote>The OpenWhisk platform supports a programming model in which developers write functional logic (called <a href="https://github.com/apache/incubator-openwhisk/blob/master/docs/actions.md#openwhisk-actions"><strong>Actions</strong></a>), in any supported programming language, that can be dynamically scheduled and run in response to associated events (via <a href="https://github.com/apache/incubator-openwhisk/blob/master/docs/triggers_rules.md#creating-triggers-and-rules"><strong>Triggers</strong></a>) from external sources ( <a href="https://github.com/apache/incubator-openwhisk/blob/master/docs/feeds.md#implementing-feeds"><strong>Feeds</strong></a>) or from HTTP requests.</blockquote><p>Using such a framework enables you to</p><ol><li>Write multiple functions in parallel increasing the effectiveness of the team.</li><li>Use multiple languages ensuring everybody is able to participate to the best of their ability.</li></ol><p>If you are new to OpenWhisk architecture, let’s do a quick crash course crash💥!</p><ul><li>Action → your code that runs on the serverless platform</li><li>Trigger → fired when an external event occurs</li><li>Rule → connects an trigger to an action</li><li>Package → bundle or modularize your actions, triggers and rules and share with other code</li></ul><p>Let’s recreate this simple example provided by <a href="https://medium.com/u/604ac4ebfc5f">James Thomas</a> in his fabulous <a href="https://github.com/IBM-Cloud/openwhisk-workshops/tree/master/bootcamp?cm_mmc=OSocial_Blog-_-Developer_IBM+Developer-_-WW_WW-_-OInfluencer-Medium-USL-openwhisk-workshop-sequence&amp;cm_mmca1=000037FD&amp;cm_mmca2=10010797">tutorial/bootcamp on OpenWhisk</a>. <a href="https://github.com/IBM-Cloud/openwhisk-workshops/tree/master/bootcamp?cm_mmc=OSocial_Blog-_-Developer_IBM+Developer-_-WW_WW-_-OInfluencer-Medium-USL-openwhisk-workshop-sequence&amp;cm_mmca1=000037FD&amp;cm_mmca2=10010797">Exercise 1.2</a> provides an example of action sequence, which simply is an action calling another action, which in turn calls a third action. The results of one action are passed onto the next one. If there is an error, the whole sequence errors out. Instead of implementing the whole sequence in a single language, we will use Javascript, Python and Swift! How cool is that! Here are are requirements</p><ol><li>Split: should take in a string and return an array of words split by space. We will write this function in <strong>Javascript</strong>.</li><li>Reverse: should take in an array and return a new array that is reverse of the input array. We will write this function in <strong>Swift</strong>.</li><li>Join: takes in an array and returns a new array that is the space delimited join of the elements. We will write this function in <strong>Python</strong>.</li></ol><p>For example, if I pass in “hello world” to this sequence, I will get back “world, hello”. What do these functions looks like?</p><h4>Create Actions</h4><p><strong>splitjs Javascript action</strong></p><pre>function main(params) {<br> var text = params.text || &quot;&quot;;<br> var words = text.split(&#39; &#39;);<br> return {words: words}<br>}</pre><p><strong>reverseswift Swift action</strong></p><pre>struct Input: Codable {</pre><pre>var words: [String]?</pre><pre>}</pre><pre>struct Output: Codable {</pre><pre>let message: [String]</pre><pre>}</pre><pre>func <strong>main</strong>(param: Input<em>,</em> completion: (Output?<em>,</em> Error?) -&gt; Void) -&gt; Void {</pre><pre>var result = <strong>Output</strong>(message: [] )</pre><pre>if let words = param.words{</pre><pre>result = <strong>Output</strong>(message: words.<strong>reversed</strong>() )</pre><pre>}</pre><pre><strong>completion</strong>(result<em>,</em> nil)</pre><pre>}</pre><p><strong>joinpy Python action</strong></p><pre>import sys</pre><pre>import json</pre><pre>def <strong>main</strong>(dict):</pre><pre><strong>print</strong>(&#39;dict &#39; + json.dumps(dict))</pre><pre>if &quot;message&quot; in dict:</pre><pre>   return {&#39;message&#39;: &#39;, &#39;.join(dict[&#39;message&#39;])}</pre><pre>else:</pre><pre>   return { &#39;message&#39;: [] }</pre><p>That’s it! The following three lines <strong><em>deploy</em></strong> these code snippets as actions on <em>openwhisk.</em></p><pre>wsk action create splitjs splitjs.js<br>wsk action create reverseswift reverseswift.swift<br>wsk action create joinpy joinpy.py</pre><p>Let’s test these out now.</p><h4>Testing our actions individually</h4><p><strong>Test split javascript action</strong></p><pre>$ wsk action invoke splitjs -r -p text &quot;hello world&quot;</pre><pre><strong><em>{<br>    &quot;words&quot;: [<br>        &quot;hello&quot;,<br>        &quot;world&quot;<br>    ]<br>}</em></strong></pre><p><strong>Test reverse swift action</strong></p><pre>$ wsk action invoke reverseswift -r -p words &#39;[&quot;hello&quot;,&quot;world&quot;]&#39;<br><strong><em>{<br>    &quot;message&quot;: [<br>        &quot;world&quot;,<br>        &quot;hello&quot;<br>    ]<br>}</em></strong></pre><p><strong>Test join python action</strong></p><pre>$ wsk action invoke joinpy -r -p message &#39;[&quot;world&quot;, &quot;hello&quot;]&#39;<br><strong><em>{<br>    &quot;message&quot;: &quot;world, hello&quot;<br>}</em></strong></pre><h4>Creating a sequence</h4><p>We can create a sequence</p><p>splitjs → reverseswift → joinpy</p><pre>wsk action create split-reverse-join  --sequence splitjs,reverseswift,joinpy</pre><p>Finally, let’s test it out!</p><pre>$ wsk action invoke split-reverse-join -r -p text &quot;Hello World&quot;<br><strong><em>{<br>    &quot;message&quot;: &quot;World, Hello&quot;<br>}</em></strong></pre><p>How cool is that! Let’s all celebrate with a Chandler dance!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*QCks4cIzllXVqUltOeecpw.gif" /></figure><p>If you are around the bay area, join me for a <a href="https://www.meetup.com/IBM-Developer-SF-Bay-Area-Meetup/events/263172151/">hands-on workshop on OpenWhisk</a> on 08/08/19. You <a href="https://www.meetup.com/IBM-Developer-SF-Bay-Area-Meetup/events/263172151/">can register here</a>. <a href="https://medium.com/u/262975298e3a">IBM Developer</a> and the API wizard <a href="https://medium.com/u/bf01a11701fe">Max Katz</a> will be around.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa38a7844926" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[3 cents for junior developers/data scientists looking for a job]]></title>
            <link>https://medium.com/@lidderupk/3-cents-for-junior-developers-data-scientists-looking-for-a-job-9f93b1c859f?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/9f93b1c859f</guid>
            <category><![CDATA[life-lessons]]></category>
            <category><![CDATA[careers]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Mon, 15 Jul 2019 00:51:52 GMT</pubDate>
            <atom:updated>2019-07-15T00:51:52.907Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/667/1*J40PyNdDSXEjQIzeOfK1PA@2x.jpeg" /></figure><p>We get some wonderful interns at IBM every year. These smart and talented young minds range from high school to college. My advise to them is threefold …</p><ol><li><strong>try new things and learn as much as you can</strong>. Not to say the learning ever stops. You are always learning in the this career, whether you are a software developer or a data scientist. Whether you are a year in or ten years in. However, now is the time to learn fearlessly, without hesitation. You have more time then you will ever have again. Make mistakes early! And you cannot make mistakes without trying things out! Oh, while you are at it, HAVE FUN!!</li><li><strong>build a portfolio</strong>. Turn everything you learn into a Github project (while being respectful of company IP). You are learning anyways, use it to your advantage. You will be surprised what projects you rely on years after you have moved on to other things. Companies are less interested in what degrees and certificates you have and more interested in what you have built so far. Do you have an app on the app store? Do you have a glitch site you can show? What about a freecodecamp portfolio?</li><li><strong>ask for help and make friends</strong>. It is hard to find mentors, but if you are lucky enough to find someone, make use of them. I mentor two people at IBM and often I learn more from them than they from me! So trust me, you are doing your mentors a favor! You know what else you are doing? You are building a network. I cannot emphasize enough how important this network is going to be for you.</li></ol><p>Alright, so you finished your internship. You went back to school and passed with flying colors! You are now looking for a job. If you ever asked me for advise and actually followed it, you have some projects, you have learned lots of things from your mentors and you have an amazing network you can ask for help. What next? How do you get help? The one mistake I see a lot of recent grads make is write an email or a message on the lines of …</p><blockquote>Hello dear mentor! I just graduated. Here is my Github profile. I am looking for a job. Do you have anything?</blockquote><p>As a mentor, I am going to put this at the bottom of my TODOs. Seems like a lot of work to look for open opportunities for you, even though I think you are a brilliant developer. It really does not motivate your mentor to help, unless your mentor is your family. Family almost always helps! The trick here to help your network help you. Make it EASY for them. Here is what I propose …</p><h4>I. Meet in person or over the phone, but be very respectful of their time</h4><p>I might be old school to think that meeting in person over coffee or speaking over the phone instead of communicating via email is more effective. An email gets lost or goes to the bottom of the pile. A face and a conversation stays in memory. I know it’s not always possible to get a face to face, but even a facetime or a zoom meeting is worth while!</p><p>Ask for only the time you need. I start with a 15 minute meeting and it is amazing what you can get done if there is a time constraint. Be respectful of their time! I have tried below before and it has worked great for me.</p><blockquote>Hello, I got your contact information from xyz. I am looking for an opportunity with your team. I have attached my resume with this email. I would appreciate your time for a small chat. I have sent you an invite for 15 minutes next Monday. I am happy to reschedule as per your convenience.</blockquote><p>Once you have a meeting arranged, have an agenda. Yes, a 15 minute meeting also need to be planned and thought of in advance. Finally, make sure your ask is clear at the end of the meeting. An ask could be to get some clarification on a role, to get an interview with a team lead or to make some connections for you, introduce you to the right person.</p><h4>II. Go to them with a job posting</h4><p>Better yet, if you see an open job posting in your network, approach them with the job posting in hand. Remember, the idea is to make it as easy as possible for your contact to get you an interview at their company/team/department.</p><blockquote>Hello, I got your contact information from xyz. I am attaching an open position (id-xyz) that I found on your career site and I believe myself to be a good fit for this role. I have attached my resume with this email. I would appreciate your time for a small chat. I have sent you an invite for 15 minutes next Monday. I am happy to reschedule as per your convenience.</blockquote><h4>III. Do your homework</h4><p>This is a big one as well. Talk with people in the position you want to join in. Ask for organization charts. Ask for pain points and strategy. Get to know your future role as intimately as possible. Better yet, create a few notes or slides on why you are a good fit for this role and potentially start off your 15 minutes meeting with a presentation. I say potentially, because every meeting is different. You can also go through your notes without actually showing them as a slide. It does not have to be an official presentation to get your point across. This will do two things for you</p><ol><li>It will impress your contact and/or mentor that you did your hw.</li><li>They are not handholding you anymore. You are again making it easier for them to help you. You are empowering them to fight for you.</li></ol><p>Finally, keep in touch with your network and mentors even when not looking for a job!</p><p>What are some of your advise for developers and designers new to IT or data science and looking for a job? Calling out some senior devs and mentors who I look up to at IBM— <a href="https://medium.com/u/52ec2bf60ec1">John Walicki </a>, <a href="https://medium.com/u/bf01a11701fe">Max Katz </a>, <a href="https://medium.com/u/a66cd0d81fb9">Gabriela de Queiroz </a>, <a href="https://medium.com/u/f19d3611f01e">Kim Newman </a>, <a href="https://medium.com/u/25bb50ccfd0c">Spencer Krum</a> , <a href="https://medium.com/u/17d50d1e1c99">Ryan Anderson</a> , <a href="https://medium.com/u/5fc0831eb57b">Vijay Bommireddipalli</a> and many more who I could not find on Medium :(</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9f93b1c859f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Postman is a delight to demo APIs and services ! My top 3 features   ✨]]></title>
            <link>https://medium.com/@lidderupk/postman-is-a-delight-to-demo-apis-and-services-my-top-3-features-fd9662875d38?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd9662875d38</guid>
            <category><![CDATA[postman]]></category>
            <category><![CDATA[api]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Fri, 12 Jul 2019 14:11:02 GMT</pubDate>
            <atom:updated>2019-07-12T14:11:02.664Z</atom:updated>
            <content:encoded><![CDATA[<h3>Postman is a delight to demo APIs and services ! My top 3 features 💫 🌟 ✨</h3><p>I use Postman extensively for most of my workshops and demos. If you haven’t heard of <a href="https://www.getpostman.com/">Postman</a>, it is an API development and management tool. I however use it as a <strong>client to demo external APIs including IBM Cognitive APIs</strong>.</p><p>You can see me using it to <a href="https://www.crowdcast.io/e/introduction-to-apache">demonstrate different features of CouchDB</a> in this <a href="https://www.crowdcast.io/e/introduction-to-apache">webinar</a>. In fact, most of the workshop was built on top of Postman. I really care about making the workshop straightforward with two goals in mind …</p><ol><li>Is everybody able to finish the workshop within the designated time?</li><li>If they are not, can I provide the end solution that they can import and learn something from?</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/925/1*sbJGngPWbsAb507GkL13XA.png" /><figcaption>Introduction to CouchDB on IBM Cloud Webinar</figcaption></figure><p>I use the following features of Postman heavily to make my life easier …</p><h4>Environment variables</h4><p>The environment variables help in two ways</p><ol><li>I set them up on the folder level so that all the requests in that folder can reuse the variables.</li><li>The participants generally have to use their own API key and this makes it easier to setup in one place.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AbJF5OPPzQW2VACP-DaU2g.png" /><figcaption>Environment variables in POSTMAN</figcaption></figure><h4>Repeat REST calls</h4><p>There is a cool feature to run REST calls multiple times. You can even put a Delay between the calls. I am running the document-create call 10 times with a delay of 2ms between calls in the example below. This will create 10 documents in CouchDB. The other way to do this would be to write a script with the 10 POST calls, or use the CouchDB SDK in the language of your choice. All good options, but this seems the simplest for a webinar/tutorial.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E5Fgm60jNG4tyk2CQWjJzQ.png" /><figcaption>Collection Runner</figcaption></figure><h4>Collections/Folders and Export/Import</h4><p>This is the coolest feature. Postman has a concept of <strong>Collections</strong> and <strong>Folders</strong> within those collections. You can then package REST calls in different folders</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/473/1*ZZORs75nnmjStjcXDcSHeg.png" /><figcaption>Collections and Folders in POSTMAN</figcaption></figure><p>I also use their export/import feature to split my workshop into multiple phases. So as to cater for all levels of participants. It also gives a good and natural break point if it is a longer workshop. I have seen some folks use github branches and tags to do the same thing.</p><p>So what are some of the things you use Postman for? Do you have other tools you use for tutorials and workshops?</p><p>Thank you <a href="https://medium.com/u/bf01a11701fe">Max Katz</a> and <a href="https://medium.com/u/262975298e3a">IBM Developer</a> for hosting the webinar on CouchDB.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd9662875d38" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[NLP, Visual Recognition and Serverless example — a chatbot moderator]]></title>
            <link>https://medium.com/@lidderupk/nlp-visual-recognition-and-serverless-example-a-chatbot-moderator-321299733b4?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/321299733b4</guid>
            <category><![CDATA[visual-recognition]]></category>
            <category><![CDATA[ibm]]></category>
            <category><![CDATA[watson]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[nlp]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Thu, 11 Jul 2019 00:25:42 GMT</pubDate>
            <atom:updated>2019-07-11T00:25:42.936Z</atom:updated>
            <content:encoded><![CDATA[<h3>NLP, Visual Recognition and Serverless example — a chatbot moderator</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/822/1*Nuawfho5sJsKpxiCoDuoKg.png" /><figcaption>IBM team presenting a webinar</figcaption></figure><p>Lisa Jung and I presented a webinar on creating a chatbot moderator that tells people to be polite when they say something rude on Slack or post an appropriate picture. For the image bit, we changed the code to treat any dog pictures as rude pictures instead of looking for explicit pictures.</p><p>We used a couple of different IBM Services in this demo.</p><ol><li>We used <a href="https://ibm.biz/BdzmSV"><strong>IBM Natural Language Understanding</strong></a> to analyze text for <strong>rudeness</strong>. You can find more information <a href="https://ibm.biz/BdzmSV">here</a>.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5imqEiN5v4MgNGcBh4xqyg.png" /><figcaption>Analyzing a speech at the Federal Reserve Bank using IBM Natural Language Understanding service</figcaption></figure><p>2. We used <a href="https://ibm.biz/BdzmSJ"><strong>IBM Visual Recognition service</strong></a> to detect explicit pictures (pictures of a dog in this case). Read more <a href="https://ibm.biz/BdzmSJ">here</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mvjfD-qKuFOMak_Vu4Hy9w.png" /><figcaption>Analyzing an image with IBM Watson Visual Recognition Service</figcaption></figure><p>3. Finally, we used <a href="https://ibm.biz/BdzmSA"><strong>IBM Cloud Functions</strong></a> as the serverless platform to glue everything together. Learn more <a href="https://ibm.biz/BdzmSA">here</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t1yS6L1qO-R836jW2DnWIw.png" /></figure><p>This is the flow of the use case:</p><ol><li>The user enters text or an image in a Slack channel.</li><li>Slack pushes this content to the IBM Cloud Function action.</li><li>If the input was text, the action invokes Watson Natural Language Understanding service to determine the rudeness.</li><li>If the input was an image, the action invokes Watson Visual Recognition service to determine if it is not safe for the channel.</li><li>If the input was “bad”, the action removed the image and tells the user to be more polite. It does so by posting directly in the Slack channel.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*15O6_eBtdHoxyhq81UGq8A.png" /><figcaption>a chatbot moderator</figcaption></figure><p>You can view the <a href="https://www.crowdcast.io/e/a-moderator-using-nlp-vr">full recording on crowdcast.io</a> [<a href="https://www.crowdcast.io/e/a-moderator-using-nlp-vr">https://www.crowdcast.io/e/a-moderator-using-nlp-vr</a>]. You can get the complete code and try it out for yourself in <a href="https://ibm.biz/BdzmSu"><strong>this github repository</strong></a>. Hope that was fun! Join us for future webinars, in person meetings and workshops!</p><p>Thanks <a href="https://medium.com/u/bf01a11701fe">Max Katz</a> and <a href="https://medium.com/u/262975298e3a">IBM Developer</a> for hosting!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=321299733b4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Overview of NLP services on IBM Cloud]]></title>
            <link>https://medium.com/@lidderupk/overview-of-nlp-services-on-ibm-cloud-b7dea11b3113?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/b7dea11b3113</guid>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Tue, 25 Jun 2019 23:11:47 GMT</pubDate>
            <atom:updated>2019-06-26T00:46:12.693Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kozFmCaoSBC6gCunvSK2fQ.png" /></figure><p>I had so much fun presenting NLP on IBM Cloud to a meetup at hackerdojo in Santa Clara. We went through creating a custom Natural Language Classifier on IBM Cloud that differentiates between HAM and SPAM emails. Classic problem that has been solved numerous times, but gives a good understanding of the space. Why is the first sentence more hammy then the second?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Etrub4NL2l9s8Fh97-rqNA.png" /><figcaption>Ham or Spam?</figcaption></figure><p>You can find the slides and workshop here: <a href="https://github.com/lidderupk/hackerdojo-nlp">https://github.com/lidderupk/hackerdojo-nlp</a></p><p>Here are some of the NLP services on IBM Cloud …</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*y5SPDnufzCVEVp3H.png" /><figcaption>NLP on IBM Cloud</figcaption></figure><p>Here is a brief overview of the different NLP services …</p><h4><a href="https://ibm.biz/BdzM8v">Watson Assistant</a></h4><p>Watson Assistant lets you build conversational interfaces into any application, device, or channel. It combines machine learning, natural language understanding, and integrated dialog tools to create conversation flows between your apps and your users.</p><ul><li><a href="https://ibm.biz/BdzM8m">Watson Assistant API</a></li><li><a href="https://ibm.biz/BdzM8a">Watson Assistant Docs</a></li><li><a href="https://ibm.biz/BdzM8G">Code Patterns</a></li></ul><h4><a href="https://ibm.biz/BdzM88">Natural Language Classifier</a></h4><p>Lets you quickly build custom classifiers for textual data.</p><ul><li><a href="https://ibm.biz/BdzM8n">NLC API</a></li><li><a href="https://ibm.biz/BdzM8e">NLC Docs</a></li><li><a href="https://ibm.biz/BdzM8b">NLC Code Patterns</a></li><li><a href="https://ibm.biz/BdzM8p">NLC Demo</a></li></ul><h4><a href="https://ibm.biz/BdzM8g">Natural Language Understanding</a></h4><p>Helps you extract entities, concepts and relationships between them in text data. The service also performs sentiment analysis giving back emotions and emotions. Additionally, you can create a custom model for some APIs to get specific results that are tailored to your domain.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QY5q1hBbxSqaTjkBDitWCA.png" /></figure><ul><li><a href="https://ibm.biz/BdzM8h">NLU API</a></li><li><a href="https://ibm.biz/BdzM8V">NLU Docs</a></li><li><a href="https://ibm.biz/BdzM8J">NLU Code Patterns</a></li><li><a href="https://ibm.biz/BdzM8A">NLU Demo</a></li></ul><h4><a href="https://ibm.biz/BdzM8L">Text to Speech(tts)</a> and <a href="https://ibm.biz/BdzM83">Speech to Text(stt)</a></h4><p>As the names suggest, these services let you convert text to speech and vice versa. Additionally, they offer a way to train these models on your domain data. That is really powerful. Stay tuned for a post with a demo.</p><ul><li><a href="https://ibm.biz/BdzM8w">STT API</a></li><li><a href="https://ibm.biz/BdzM8k">STT Docs</a></li><li><a href="https://ibm.biz/BdzM8T">STT Demo</a></li><li><a href="https://ibm.biz/BdzM8C">TTS API</a></li><li><a href="https://ibm.biz/BdzM8Q">TTS Docs</a></li><li><a href="https://ibm.biz/BdzM89">TTS Demo</a></li></ul><p>These are some of the black box NLP services on IBM Cloud. They can however be extended and trained on your custom domain data. Additionally, if you have coded a model from scratch using a ML framework like scikit-learn or a DL framework like Keras, you can deploy it on IBM cloud using Watson Machine Learning (WML).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FpSPw9nv2HlmDmmhvVGu7g.png" /></figure><blockquote>So … until next time internet peeps!</blockquote><p>Thanks <a href="https://medium.com/u/bf01a11701fe">Max Katz</a> and <a href="https://medium.com/u/262975298e3a">IBM Developer</a> for providing space and delicious pizza at the meetup.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b7dea11b3113" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introducing MAX— Model Asset Exchange]]></title>
            <link>https://medium.com/@lidderupk/introducing-max-model-asset-exchange-e437970bfc8f?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/e437970bfc8f</guid>
            <category><![CDATA[ibm]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[model-asset-exchange]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Mon, 17 Jun 2019 17:12:04 GMT</pubDate>
            <atom:updated>2019-06-25T22:45:40.216Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E99n0jgH4eL9rpfyFELN5Q.png" /></figure><p><strong><em>Data Science is about learning</em></strong>. I don’t mean supervised learning, unsupervised learning, reinforcement learning or transfer learning! I mean <strong><em>LEARNING</em></strong>! From books. From code. From your colleagues and friends. From stackoverflow! You are constantly learning and improving your skill set. New models and breakthroughs are being announced at a faster rate than ever. As data scientists, you need to be aware of what is going on in this field and stay on top of the latest developments. IBM has a few open source projects that I want to write about that will help you keep up to date. We will look MAX or Model Asset Exchange in this post.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/490/1*ytBdCoJZcGBZ-igTzjvdsw.gif" /><figcaption>Learning</figcaption></figure><p>MAX is a repository of open source Machine Learning and Deep Learning models developed by my colleagues and friends at IBM. They include topics ranging from object recognition, pose detector, audio analysis and age estimator. The team puts out a new model every now and then. So what’s the big deal ?</p><blockquote>You can clone, fork, improve upon, provide feedback and learn from these models! You LEARN! You learn how to make one yourself and make it usable! You learn why it works well. You learn why some it does not work well and more importantly why it does not work with certain inputs.</blockquote><p>My dear friend and colleague <a href="https://medium.com/u/df9e907db0b1">Patrick Titzler</a> gave a talk in San Francisco recently on MAX and showed us how to use Node-RED to quickly test some of these models. I met <a href="https://www.instagram.com/throughshutter/">Ayush</a> at the same meetup. Ayush is using two models from MAX to facilitate his everyday hobby — photography. I will save that for another post later. Let’s look at my favorite MAX models …</p><h4><a href="https://developer.ibm.com/exchanges/models/all/max-object-detector?cm_mmc=OSocial_Blog-_-Developer_IBM+Developer-_-WW_WW-_-ibmdev-OInfluencer-Medium-USL-ibm-max-object-detector&amp;cm_mmca1=000037FD&amp;cm_mmca2=10010797">Object Detector</a></h4><p>This model recognizes the objects present in an image from the 80 different high-level classes of objects in the COCO Dataset. This problem has been solved for most part and if the framework/platform you are using does not do well with the default results, try creating a custom classifier (insert url) with Watson Visual Recognition. As an example, I gave the leading visual recognition APIs an image of Raspberry Pi and they mostly came back with “computer chip”, “mother board” or “electronic”. Those are all amazing results, but I wanted more. I was able to train a custom classifier using Watson to tell me with high confidence if the picture was “Raspberry Pi Model A” or “Raspberry Pi Model B” 😳. Pretty cool stuff! You can read about the customer classifier in this <a href="https://medium.com/coinmonks/watson-visual-recognition-custom-models-9cbcc20a6f98">medium post</a>. That being said, this open source model still makes for a great learning resource. You can also try these patterns in Node-RED, more coming on this option in the next post! Pretty cool that it got the bicycle just with what was visible in the picture! The bicycle is hanging upside down in my room 🆒!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VIdcctKY4glS0IEWGMQJ4g.png" /><figcaption>Node-RED for MAX</figcaption></figure><h4><a href="https://developer.ibm.com/exchanges/models/all/max-facial-age-estimator?cm_mmc=OSocial_Blog-_-Developer_IBM+Developer-_-WW_WW-_-ibmdev-OInfluencer-Medium-USL-ibm-max-age-estimator&amp;cm_mmca1=000037FD&amp;cm_mmca2=10010797">Facial Age Estimator</a></h4><p>This model detects faces in an image, extracts facial features for each face detected and finally predicts the age of each face.</p><p>This is my favorite by far because it gets me younger every time :). At first, I was quite annoyed that IBM would put out a model that did not do as well as I expected it to. Sometimes I shave and other times I have a stubble. It gives me a different result in day light vs inside a dimly lit conference hall. But then I realized that these are all learning opportunities for me. Questions that we need to ask of each model. I started looking at the training data and quickly realized that unless I am in Hollywood and have five lbs of makeup on, it wont be able to guess my age. The training data is IMDB imageset that you can <a href="https://developer.ibm.com/exchanges/models/all/max-facial-age-estimator?cm_mmc=OSocial_Blog-_-Developer_IBM+Developer-_-WW_WW-_-ibmdev-OInfluencer-Medium-USL-ibm-max-age-estimator&amp;cm_mmca1=000037FD&amp;cm_mmca2=10010797">find here</a>. We will look at this dataset closely in an upcoming blog post.</p><p><a href="https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/">IMDB-WIKI - 500k+ face images with age and gender labels</a></p><p>We Now have an<a href="https://developer.ibm.com/patterns/estimate-ages-for-detected-human-faces?cm_mmc=OSocial_Blog-_-Developer_IBM+Developer-_-WW_WW-_-ibmdev-OInfluencer-Medium-USL-max-estimate-age&amp;cm_mmca1=000037FD&amp;cm_mmca2=10010797"> IBM code pattern</a> that will help you create a web application that uses this model! These world leaders are awfully young!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nos_L2nOT37mr1snpxLh7Q.png" /><figcaption>G-7 leaders</figcaption></figure><p>It does fairly well with my image, only off by two … I will let you guess in which direction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8u0DswIGrLub0HC_wQEqJQ.png" /><figcaption>Upkar Lidder</figcaption></figure><h4><a href="https://developer.ibm.com/exchanges/models/all/max-image-caption-generator?cm_mmc=OSocial_Blog-_-Developer_IBM+Developer-_-WW_WW-_-ibmdev-OInfluencer-Medium-USL-ibm-max-image-caption&amp;cm_mmca1=000037FD&amp;cm_mmca2=10010797">Image Caption Generator</a></h4><p>This is the model Ayush was using to generate captions for his Instagram account. By the way, you can follow him here: <a href="https://www.instagram.com/throughshutter/">https://www.instagram.com/throughshutter</a></p><p>This model generates captions from a fixed vocabulary that describe the contents of images in the COCO Dataset. Even though the training dataset is limited (80 object categories), it does pretty well. Here is an example</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/950/1*SqfWTFfphEaOn2SjU4Y4AA.png" /><figcaption><strong><em>a man in a suit talking on a cell phone</em></strong></figcaption></figure><blockquote><strong><em>A man in a suit talking on a cell phone</em></strong></blockquote><p>The caption says <strong><em>a man in a suit talking on a cell phone. </em></strong>That is not bad at all! There are plenty of other models you can try on MAX. But let’s look at how you can use Node-RED to quickly test a few of these. The MAX team has created some assets to help us out. If you have never used Node-RED before, get ready to fall in love 💕.</p><h4><a href="https://ibm.biz/BdzPh9">Audio Classifier</a></h4><p>This one is so much fun. <em>This model recognizes a signed 16-bit PCM wav file as an input, generates embeddings, applies </em><a href="https://github.com/tensorflow/models/tree/master/research/audioset#output-embeddings"><em>PCA transformation/quantization</em></a><em>, uses the embeddings as an input to a multi-attention classifier and outputs top 5 class predictions and probabilities as output.</em> WHAT 🤯!! Basically, it identifies sounds and noise. The data is coming from <a href="https://research.google.com/audioset/ontology/index.html">AudioSet</a>. I am building some cool demos and games to use this model. Stay tuned!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*m7St2oLWP1Gm2HuZoOPtQg.png" /></figure><h4><a href="https://ibm.biz/BdzPhC">Human Pose Estimator</a></h4><p>I had a lot of fun with this one at a recent data science workshop! This model detects humans and their poses in a given image. The model first detects the humans in the input image and then identifies the body parts, including nose, neck, eyes, shoulders, elbows, wrists, hips, knees, and ankles. Next, each pair of associated body parts is connected by a pose line.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x69Jto9UfnwN9MFk_fnCzw.png" /><figcaption>Human Pose Estimator</figcaption></figure><p>Those are some of my favorite models. There are lots more in the Model Asset Exchange. <a href="https://ibm.biz/BdzPhV">Give them a try and let me know which ones you like</a>! I have some more posts coming up on how to use/deploy <a href="https://ibm.biz/BdzPhV">some of these models</a>.</p><p><a href="https://medium.com/u/bf01a11701fe">Max Katz</a>, sorry for the confusion in advance :). I will write a separate post to introduce you! <a href="https://medium.com/u/df9e907db0b1">Patrick Titzler</a>, thank you for the wonderful meetup! <a href="https://medium.com/u/a66cd0d81fb9">Gabriela de Queiroz</a> and Simon, thank you for demoing MAX at the <a href="https://aiml-developer-summit.splashthat.com/">IBM AI-ML Summit</a> in San Francisco.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e437970bfc8f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Another simplest serverless function ‍♀ — but seriously ]]></title>
            <link>https://medium.com/@lidderupk/another-simplest-serverless-function-but-seriously-3223b4ebcb6d?source=rss-50c02028dfb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/3223b4ebcb6d</guid>
            <category><![CDATA[serverless]]></category>
            <category><![CDATA[ibmcloudfunctions]]></category>
            <category><![CDATA[ibm]]></category>
            <dc:creator><![CDATA[Upkar Lidder]]></dc:creator>
            <pubDate>Thu, 13 Jun 2019 00:54:19 GMT</pubDate>
            <atom:updated>2019-06-13T00:54:19.833Z</atom:updated>
            <content:encoded><![CDATA[<h3>Another simplest serverless function 🤷‍♀ — but seriously 💖</h3><p>I have come across a couple of posts with the title <strong><em>simplest serverless functions</em></strong> and my initial thoughts are …</p><ol><li>It is not that simple if you have to write up a config file to post a <strong><em>hello world</em></strong> function 🤦‍♂ !</li><li>How easy is it to now extend this function to something more usable ?</li></ol><p>So here we go again … this example is based on <strong><em>Apache OpenWhisk</em></strong>, <strong><em>an open source distributed serverless computing platform</em></strong>. However, a serverless system on your desktop/laptop is no different than spinning up a nodejs express app or a flask app.<strong><em> You are missing out on scale, performance and distributed nature of OpenWhisk</em></strong>. In order to make it even simpler, we will use IBM Cloud Functions, a managed version of Apache OpenWhisk.</p><h4>Check this out …</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ETrPUMm2aKCbzsOMhI8CEA.gif" /></figure><p>In the next few posts, we will learn to extend this function to do something more usable. But wasn’t that easy !</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*OJvehEDxKyjCMzvxsdiaMg.gif" /></figure><p>See you next time ! In the meantime, read up on some <a href="https://ibm.biz/BdzPiS">serverless code patterns</a>.</p><p><a href="https://medium.com/u/bf01a11701fe">Max Katz</a>, thanks for this post idea !</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3223b4ebcb6d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>