<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Anish Shirodkar</title>
    <description>The latest articles on DEV Community by Anish Shirodkar (@anishs).</description>
    <link>https://dev.to/anishs</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4248529%2F9f56a28a-f819-4f33-8501-c0fe42313618.JPG</url>
      <title>DEV Community: Anish Shirodkar</title>
      <link>https://dev.to/anishs</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/anishs"/>
    <language>en</language>
    <item>
      <title>Oxyflame: How We Mapped Where NASA's Space Fire Evidence Ends (Best Use of Science, NASA Space Apps NYC 2026)</title>
      <dc:creator>Anish Shirodkar</dc:creator>
      <pubDate>Wed, 18 Nov 2026 02:45:01 +0000</pubDate>
      <link>https://dev.to/anishs/oxyflame-how-we-mapped-where-nasas-space-fire-evidence-ends-best-use-of-science-nasa-space-apps-nj2</link>
      <guid>https://dev.to/anishs/oxyflame-how-we-mapped-where-nasas-space-fire-evidence-ends-best-use-of-science-nasa-space-apps-nj2</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/mlh-hackathon"&gt;MLH x DEV Writing Challenge&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Oxyflame&lt;/strong&gt; asks a simple question: &lt;em&gt;how does fire behave in space, and where does our evidence run out?&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;NASA has flown dozens of combustion experiments on the ISS. Astronauts set fabrics, droplets and gases on fire inside sealed boxes and recorded how the flames spread and went out. The data is public in NASA's Physical Sciences Informatics (PSI) repository, but it lives in test matrices, PDFs and papers spread across separate investigation pages. If you're designing a Moon habitat and want to know whether anyone has tested a material in conditions like yours, there's no easy way to find out.&lt;/p&gt;

&lt;p&gt;Oxyflame pulls &lt;strong&gt;24 PSI combustion investigations&lt;/strong&gt; (19 ISS flight experiments and 5 ground studies built on their data) into one comparable dataset and puts a mission-aware dashboard on top of it:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Pick an environment.&lt;/strong&gt; Earth, ISS, the Moon challenge scenario (30% O₂ at 10.2 psi), a Moon habitat, a rover, Mars or a spacesuit. Oxyflame ranks the NASA experiments that tested conditions closest to it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;See the evidence strength, not a safety rating.&lt;/strong&gt; Each answer shows how well the existing tests cover your scenario and names the gap honestly. For example, no tested run in our dataset matches lunar gravity.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ask questions in plain English.&lt;/strong&gt; Every answer cites the PSI investigation it came from, like &lt;code&gt;[PSI-26]&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Find the missing experiment.&lt;/strong&gt; The app points to the conditions nobody in the dataset has tested yet.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Text it.&lt;/strong&gt; No internet? Send "MOON fabric flammability" to a phone number and get the top matching investigation back by SMS.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The biggest lesson was about honesty in AI tools. An AI that confidently says "this fabric is safe on the Moon" is dangerous. So we made a rule: Oxyflame never says a material is safe, and gaps are described as "not in our curated dataset," never "never studied."&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;🌐 Website: &lt;a href="https://oxyflame.oxyflame.workers.dev" rel="noopener noreferrer"&gt;https://oxyflame.oxyflame.workers.dev&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;📱 Installable mobile app: &lt;a href="https://oxyflame.pages.dev" rel="noopener noreferrer"&gt;https://oxyflame.pages.dev&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;💻 Code: &lt;a href="https://github.com/Anish0104/SpaceApps" rel="noopener noreferrer"&gt;https://github.com/Anish0104/SpaceApps&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;🎥 Demo video: &lt;a href="https://youtu.be/C8Iw6QJeiHY" rel="noopener noreferrer"&gt;https://youtu.be/C8Iw6QJeiHY&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fa89tetj5gq8mngnr2nwx.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fa89tetj5gq8mngnr2nwx.png" alt=" " width="800" height="461"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6wfqm04tcayz2szrv3zb.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6wfqm04tcayz2szrv3zb.png" alt=" " width="800" height="461"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3f0khz2p5gkg09j5tjpx.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3f0khz2p5gkg09j5tjpx.png" alt=" " width="800" height="461"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fgdwq6tmvfyvni5q7kc1m.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fgdwq6tmvfyvni5q7kc1m.png" alt=" " width="800" height="461"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Partner Technologies
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Google Gemini&lt;/strong&gt; powers the "Ask" feature. Answers are built only from our curated PSI dataset, and each claim has to cite an investigation ID. Free-tier quota was tight over a hackathon weekend, so we added a cache and pre-warmed the answers to our demo questions. That way a live demo never depends on a quota reset.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Cloudflare&lt;/strong&gt; runs the whole stack. One Worker serves the website and the API, Cloudflare Pages hosts the installable mobile app, and Workers AI handles embeddings for hybrid (keyword + semantic) search over the investigations. Deploying was one &lt;code&gt;pnpm run deploy&lt;/code&gt;, and the free tier covered everything.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Twilio&lt;/strong&gt; powers "text Oxyflame," built by my teammate Juvana. Twilio sends incoming texts to a &lt;code&gt;/api/sms&lt;/code&gt; webhook on our Worker. The SMS replies deliberately skip Gemini and use the same rules engine and template answers as the website, so a text reply can never invent a number. Every request is checked against Twilio's signature (anything else gets a 403), and each phone number is limited to 8 texts a minute. Text "HELP" for instructions.&lt;/p&gt;

&lt;h2&gt;
  
  
  Hackathon Experience
&lt;/h2&gt;

&lt;p&gt;I competed at &lt;strong&gt;NASA Space Apps Challenge NYC&lt;/strong&gt; (Nov 13–15, 2026) with my team &lt;strong&gt;HireUs&lt;/strong&gt;: Navneet on data, Juvana on science and evaluation, and me on the AI side. We took on the &lt;strong&gt;Flame in Freefall&lt;/strong&gt; challenge.&lt;/p&gt;

&lt;p&gt;What broke at 3am (more or less):&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;The data wasn't a dataset.&lt;/strong&gt; Most of what we needed was in documents and test matrices, not CSVs, so Navneet spent the first night turning 24 investigation pages into one table.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;US carriers block texts from unregistered numbers.&lt;/strong&gt; A2P 10DLC registration takes days, so our SMS demo ran through the live, signature-checked webhook instead of a public number.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;What I'll remember most: the food was great, we got almost no sleep, and it was still one of the best weekends I've had. Running on no sleep with two teammates who kept going until the last minute is how hackathon memories get made.&lt;/p&gt;

&lt;p&gt;We won &lt;strong&gt;Best Use of Science&lt;/strong&gt; 🏆. I think we won because we treated "what the data &lt;em&gt;doesn't&lt;/em&gt; tell you" as a feature instead of hiding it.&lt;/p&gt;

&lt;p&gt;Thanks to my teammates &lt;a class="mentioned-user" href="https://dev.to/yourselfthere"&gt;@yourselfthere&lt;/a&gt; and &lt;a class="mentioned-user" href="https://dev.to/juvanak"&gt;@juvanak&lt;/a&gt; , the Space Apps NYC organizers, and NASA for keeping PSI open.&lt;/p&gt;

</description>
      <category>mlhacks</category>
      <category>devchallenge</category>
      <category>hackathon</category>
      <category>nasa</category>
    </item>
  </channel>
</rss>
