<?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 Andrew Kozlik on Medium]]></title>
        <description><![CDATA[Stories by Andrew Kozlik on Medium]]></description>
        <link>https://medium.com/@andrewkozlik?source=rss-6992a043f0fb------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*ZDOsRVPATolkDvvMEz-HDQ.jpeg</url>
            <title>Stories by Andrew Kozlik on Medium</title>
            <link>https://medium.com/@andrewkozlik?source=rss-6992a043f0fb------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 07:01:31 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@andrewkozlik/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[Finding Recipes Faster with Atlas Search]]></title>
            <link>https://medium.com/@andrewkozlik/paying-off-technical-debt-with-the-debt-snowball-be4370aef9bb?source=rss-6992a043f0fb------2</link>
            <guid isPermaLink="false">https://medium.com/p/be4370aef9bb</guid>
            <category><![CDATA[mongodb]]></category>
            <category><![CDATA[technical-debt]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[engineering]]></category>
            <dc:creator><![CDATA[Andrew Kozlik]]></dc:creator>
            <pubDate>Mon, 14 Sep 2020 17:17:13 GMT</pubDate>
            <atom:updated>2020-09-15T19:54:55.356Z</atom:updated>
            <content:encoded><![CDATA[<h4>We paid off long overdue technical debt while making RecipeBox searches more relevant for our users.</h4><p>Last night, I paid off some technical debt for <a href="http://www.getrecipebox.com/">RecipeBox</a>. We migrated our recipe search functional off of our MongoDB database and into a dedicated Atlas Search index. This change drastically reduced the search query response times against our MongoDB instance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/880/1*t3D78hWdoR7btImzXj7GOg.png" /><figcaption>Each dot is a long running DB query over 100ms</figcaption></figure><p>The verdict is still out but this change also reduced our server’s average response times.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*R8WAT88AA9my3kAGXaiAtQ.png" /><figcaption>The tall and light purple boxes are the highest response time transactions during that time period.</figcaption></figure><p>Only two of us do all the engineering for RecipeBox. One of my partners, Tom, handles our iOS development and I do everything web related. Limited resources means we have to balance between building things “right” and getting features out the door fast.</p><p>Search is a feature where we accepted technical debt to pay down later. I wanted to talk a little about how we took a snowball approach to paying that debt off.</p><p>When we first implemented search over a year ago we had around 40,000 recipes in our database. Users could search for recipes by category, description, or ingredients.</p><p>For this first iteration we did a plain text search. Users could search for “tomato” and we would return recipes with “tomato” in the one of a few fields.</p><p>This was slow, but with only 40,000 records it wasn’t unmanageable and it let us get something out the door. A search for “tomato” would return recipes that had the word “tomato”, “tomatoes”, “tomato sauce”, etc in the DB. There was no weighting, search results weren’t perfect, but they worked.</p><p>“We’ll fix it later,” we said.</p><p>By recipe 200,000 the cracks in our search implementation began to show. Queries were slow and clogging up the transaction queue on our web servers. We knew we wanted to migrate to a dedicated search platform but we were fighting bigger fires at the time. I didn’t have time or resources to invest in standing up a dedicated search service so we needed temporary solution that we could build fast.</p><p>We can speed up our database searches by adding indexes on the data we want to search. This comes at a storage cost but allows our searches to become near instant. Yet, limitations in our data prevented us from being able to index the fields we wanted to search. Unfortunately, indexing wasn’t going to be an option for us at that time.</p><p>Our long query times were stemming from doing partial text matches on our document store. A temporary solution would be to put in place an exact match mechanism that did not rely on partial text matches. This would speed up our queries but the searched data would need to exactly match a user’s query. A search for “tomato” would no longer return results containing “tomatoes” or “tomato sauce”.</p><p>We decided this was a good opportunity to try a hashing approach. We hashed our recipe URLs and ingredients and updated the records in the database. We would hash the user’s query and do a lookup in the database. Search wouldn’t be perfect, but it would be more performant.</p><p>Hashing meant we took a string like “https://www.budgetbytes.com/slow-cooker-hamburger-stew/“ and reduced it to a string like “5141c1572ef30f469d99e77c8edd2fc0”. That string is small enough for MongoDB to index.</p><p>After updating our production data we added the indexes. This change reduced our server load but now users could only search for exact matches.</p><p>“We’ll fix it later,” we said.</p><p>As we grew to 800,000 recipes we started hitting limitations on what mLab could do for us. MongoDB acquired mLab and we needed to migrate to Atlas by November. We completed the migration and found a tool called Atlas Search that solves this problem. Atlas provides the tool as part of their services so there would be no extra overhead for us. We would have a dedicated search platform that syncs with our production database and is easy to maintain.</p><p>Over a few evenings we updated our search implementation to leverage Atlas Search. Offloading the queries from our database plummeted our average query response times. This improvement also dropped our server response times since we were no longer waiting on those long queries to run.</p><p>“It’s fixed! (After we continue to watch our logs)” we said.</p><p>This was a year long journey of weighing acceptable technical debt against how that debt would affect our users. We took on that debt and chipped away at it as time, customer needs, and resourcing permitted. At times, we accepted that we may be making the user experience worse but improving the platform as a whole. This allowed us to make broader progress on our product and get more features to market faster. We also carved out the time to address the debt and get it completely paid off.</p><p>As we gear up to hit our 1,000,000th recipe I’m excited to see recipe searching become faster and more accurate. We’ll continue to take on and pay off technical debt as we grow RecipeBox.</p><p>RecipeBox is available on the <a href="http://www.getrecipebox.com/">web</a> and <a href="https://apps.apple.com/us/app/recipebox-save-your-recipes/id1433840494">iOS</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=be4370aef9bb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Impact Happens Outside of the Command Line]]></title>
            <link>https://medium.com/code-for-america/impact-happens-outside-of-the-command-line-d2c8bcd93233?source=rss-6992a043f0fb------2</link>
            <guid isPermaLink="false">https://medium.com/p/d2c8bcd93233</guid>
            <category><![CDATA[cfa-network-news]]></category>
            <category><![CDATA[civictech]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[volunteer]]></category>
            <category><![CDATA[government]]></category>
            <dc:creator><![CDATA[Andrew Kozlik]]></dc:creator>
            <pubDate>Mon, 30 Oct 2017 22:41:27 GMT</pubDate>
            <atom:updated>2017-10-30T23:01:40.911Z</atom:updated>
            <content:encoded><![CDATA[<p><em>The following post is a condensed version of the closing speech I gave at Code for America’s inaugural Brigade Congress.</em></p><p>Over the last two days I’ve spoken to leaders from 38 cities across the US and Canada. I’ve heard from cities as large as LA and New York but also as small as Portland Maine and Newport Rhode Island. We’re represented by Brigades as old as five years and as young as eight weeks. The network is growing and we’re awakening something in people that drives them to improve their cities and their country.</p><p>This weekend we’ve had existential conversations and asked ourselves challenging questions. In one of our sessions Jim Van Fleet from Charlotte said, “Technical problems seldom outweigh the people problems” and it really resonated with me. Some of the largest challenges we face are not “how do I get this to compile” (though sometimes does feel like the biggest challenge for those who use my code). Our challenges lie in identifying how we can use our skills to have the most impact in our communities. As <a href="https://twitter.com/mishmosh">Michelle Lee</a> said, “Impact happens outside of the command line.”</p><p>We’ve shared our successes, failures, and fears. We’ve asked hard questions like “how do we engage and create a more diverse community” and “how are we unintentionally excluding people from participating.”</p><p>Emma Burnett from Code for Maine turned on a light bulb in my head when she said they run not “hackathons” but “civic design fests”. Their thought was the term “hackathon” can be exclusionary because it implies a technical background is needed to participate. By using more inclusive language we can encourage others from outside the community to participate. As James Lott from Code for Philly said “It’s a civic movement, not a tech movement.” Who knows? Maybe next year we’ll be celebrating National Day of Civic Design. I love the idea of challenging how we’re operating and seeing where we can improve.</p><p>Sometimes we get caught up in the day to day grind of operating our Brigades we forget we’re part of a larger network and have other leaders facing the same challenges as us.</p><p>The power and potential of the Brigade network was really unlocked for me this September during Hurricane Irma. Code for America’s Christopher Whitaker reached out to the Florida Brigade networks and said “Hey, Sketch City in Houston told me they built this amazing tool for disaster response called HarveyNeeds.org and they deployed a version for Florida to use.” Houston rolled all of their energy from Harvey and put something together for Florida to use. Brigade leaders in Miami and Fort Lauderdale activated engineers and designers to build on top of the work while Orlando used a scheduled open hack night for data entry. Within a few days, irmaresponse.org was launched.</p><p>Think about what happened here. Sketch City built a tool to solve a problem they were facing. The same problem cropped up in Florida and they saw an opportunity. They reached out through Whitaker and made us aware someone solved a problem we were facing. Rather than building something from scratch we had the opportunity to adopt and build on top of an existing tool to meet our needs. We’re now discussing the possibility of leveraging the same tool for tornado response in Oklahoma. This would not have been possible without the Brigade network.</p><p>So as you’re taking your planes, trains, and automobiles back to your hometowns, remember the energy you felt in this building. Continue to drive your Brigades and keep your members engaged. Lead by example and push others in your Brigades to become leaders. Consider serving our network in the next National Advisory Council and bring to the table all the ways we can improve.</p><p>Remember there’s strength in numbers. Keep in touch with all those you met this weekend. Nurture those slack channels and keep them alive!</p><p>If you loved meeting in person be sure to come to <a href="http://www.govexec.com/sponsor-content/code-america-summit-2018/">Code for America Summit in Oakland from May 30th — June 2nd</a>.</p><p>Continue the conversations you’ve had this weekend and pay attention to what problems other Brigades are facing and offer to help. By working together, the Code for America Brigade Network will be louder than the crack in the bell.</p><p>Thank you.</p><p>—</p><p>Andrew Kozlik is co-captain with Erin Denton of Code for Orlando and serves on the National Advisory Council.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d2c8bcd93233" width="1" height="1" alt=""><hr><p><a href="https://medium.com/code-for-america/impact-happens-outside-of-the-command-line-d2c8bcd93233">Impact Happens Outside of the Command Line</a> was originally published in <a href="https://medium.com/code-for-america">Code for America Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Thrilled, nervous, and excited to go on our trip tomorrow.]]></title>
            <link>https://medium.com/@andrewkozlik/thrilled-nervous-and-excited-to-go-on-our-trip-tomorrow-18dfb5b42596?source=rss-6992a043f0fb------2</link>
            <guid isPermaLink="false">https://medium.com/p/18dfb5b42596</guid>
            <category><![CDATA[roadtrip]]></category>
            <category><![CDATA[camping]]></category>
            <category><![CDATA[travel]]></category>
            <dc:creator><![CDATA[Andrew Kozlik]]></dc:creator>
            <pubDate>Tue, 07 Jun 2016 03:49:16 GMT</pubDate>
            <atom:updated>2016-06-07T03:49:16.972Z</atom:updated>
            <content:encoded><![CDATA[<p>Thrilled, nervous, and excited to go on our trip tomorrow. We spent the day sorting through our gear, loading it all up in truck, and doing final checks on equipment. Tent is looking great and we have plenty of fluids in the car.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*o5OPO2D0ts7jwHOep5nbXQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*o1ouzGsc4rIYVY7oM7ItNA.jpeg" /></figure><p>Going to miss all the friends and family while we’re gone. Breaks my heart not to be able to bring dog but she’ll be much more comfortable staying with Deb and kitty cat. Two months is a long time without Harley but it’ll go by quick and we’ll be together again.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qJb1MTEqNPC78JSMARusSQ.jpeg" /></figure><p>Tomorrow morning marks the first leg of our two month trip around the US.</p><p>#freedom here we come!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=18dfb5b42596" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>