<?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: Muskan Khatoon</title>
    <description>The latest articles on DEV Community by Muskan Khatoon (@muskan110).</description>
    <link>https://dev.to/muskan110</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%2F4151623%2F533d655f-a5f3-4944-8f3b-3f9e31314f60.png</url>
      <title>DEV Community: Muskan Khatoon</title>
      <link>https://dev.to/muskan110</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/muskan110"/>
    <language>en</language>
    <item>
      <title>🦋 Sidewalk Safari: Turning Ordinary Walks into Tiny Nature Adventures</title>
      <dc:creator>Muskan Khatoon</dc:creator>
      <pubDate>Sat, 10 Oct 2026 19:48:28 +0000</pubDate>
      <link>https://dev.to/muskan110/sidewalk-safari-turning-ordinary-walks-into-tiny-nature-adventures-3k8</link>
      <guid>https://dev.to/muskan110/sidewalk-safari-turning-ordinary-walks-into-tiny-nature-adventures-3k8</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week1-2026-10-05"&gt;Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

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

&lt;p&gt;Have you ever reached class and realised you don't remember a single thing you walked past?&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Sidewalk Safari&lt;/strong&gt; turns an ordinary 5–15 minute walk into a small nature expedition. You choose how long you have, where you are (campus, park, neighborhood, anywhere outside) and how tricky you want it. A language model running on your own computer then writes three short observation missions:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;em&gt;Find one leaf that shows three different shades of green.&lt;/em&gt;&lt;/li&gt;
&lt;li&gt;&lt;em&gt;Look for a sign that an insect has visited a plant.&lt;/em&gt;&lt;/li&gt;
&lt;li&gt;&lt;em&gt;Pick a tree and watch how light and shadow change its leaves.&lt;/em&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Then you put your phone away. Pocket Mode turns the screen into a dark timer that says "Phone in your pocket. Go find your missions." You tick missions off as you find them, each one reveals a short field note about what you just noticed, and when you're done you write a few lines in a field journal that lives in your browser.&lt;/p&gt;

&lt;p&gt;It's for students between classes, anyone who wants a reason to look up, and anyone who'd like their phone to be the start of a walk rather than the whole of it. There are no accounts, no streaks, no notifications and no leaderboards. The screen should be the shortest part of the experience.&lt;/p&gt;

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

&lt;p&gt;🎥 &lt;strong&gt;Video of a real outdoor safari:&lt;/strong&gt; &lt;a href="https://drive.google.com/file/d/10rD9phaRyp_E-Me2jLPzj3cYT5CRoqpZ/view?usp=sharing" rel="noopener noreferrer"&gt;https://drive.google.com/file/d/10rD9phaRyp_E-Me2jLPzj3cYT5CRoqpZ/view?usp=sharing&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Sidewalk Safari runs on your own machine, so the demo is a video of an actual walk rather than a hosted page.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Taking it outside:&lt;/strong&gt; I took Sidewalk Safari for a short walk around my college campus, spending about 10 minutes exploring the little details I usually walk past without noticing. I completed three observation missions, and one of the most interesting discoveries was noticing how leaves on the same plant could have different shades of green and patterns. The brief loading wait felt a little noticeable at first, but once the missions were generated, I could put my phone away and focus on the world around me. It reminded me that sometimes, we don't need to travel somewhere special to discover something new we just need a reason to look closer.&lt;/p&gt;

&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/muskan-110" rel="noopener noreferrer"&gt;
        muskan-110
      &lt;/a&gt; / &lt;a href="https://github.com/muskan-110/Sidewalk-Safari" rel="noopener noreferrer"&gt;
        Sidewalk-Safari
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;🌿 Sidewalk Safari&lt;/h1&gt;
&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;Turn an ordinary 5–15 minute walk into a small nature expedition. The AI runs on your own computer.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;Sidewalk Safari uses a local open-weight language model (Gemma 3 4B via Ollama) to write three short observation missions for your walk. You put your phone away, find them in the real world, then come back and record what you noticed in a field journal saved in your browser.&lt;/p&gt;
&lt;p&gt;Built for the &lt;a href="https://dev.to/challenges/hacktoberfest-week1-2026-10-05" rel="nofollow"&gt;Hacktoberfest Open-Source AI Challenge, Week 1: Touch Grass&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://hacktoberfest.com" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/5cdf62d62f9c95086c90f00c438cc2dcc98c17fce950db6ce93944fd5e1ad74c/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4861636b746f626572666573742d323032362d6f72616e67653f6c6f676f3d6861636b746f62657266657374266c6f676f436f6c6f723d7768697465" alt="Hacktoberfest 2026"&gt;&lt;/a&gt;
&lt;a href="https://dev.to" rel="nofollow"&gt;&lt;img src="https://camo.githubusercontent.com/17f26b8c490dcf9e2f2594e1197e40ecfec44d1d307fe30bf592b65637719f42/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4445562532304368616c6c656e67652d546f75636825323047726173732d626c61636b3f6c6f676f3d646576646f74746f266c6f676f436f6c6f723d7768697465" alt="DEV Challenge"&gt;&lt;/a&gt;
&lt;a href="https://github.com/muskan-110/Sidewalk-Safari/LICENSE" rel="noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/d6bc2b26794002c24d023acaab01b6dbb953c57ab9cb80ba5b8aa2f2bd5de99a/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4c6963656e73652d4d49542d626c7565" alt="License: MIT"&gt;&lt;/a&gt;
&lt;a rel="noopener noreferrer nofollow" href="https://camo.githubusercontent.com/9b1152826e41a338cf09e1ef8128dc3ea7cd53b78b6a28dc37642baf88efdd38/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4261636b656e642d466173744150492d3030393638383f6c6f676f3d66617374617069266c6f676f436f6c6f723d7768697465"&gt;&lt;img src="https://camo.githubusercontent.com/9b1152826e41a338cf09e1ef8128dc3ea7cd53b78b6a28dc37642baf88efdd38/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4261636b656e642d466173744150492d3030393638383f6c6f676f3d66617374617069266c6f676f436f6c6f723d7768697465" alt="Backend"&gt;&lt;/a&gt;
&lt;a rel="noopener noreferrer nofollow" href="https://camo.githubusercontent.com/f2bdf5331e945f9e41e00ec42131818fc4f2ab27e8d0e320e4de2d493d0f5998/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f46726f6e74656e642d5265616374253230253242253230566974652d3230323332413f6c6f676f3d7265616374266c6f676f436f6c6f723d363144414642"&gt;&lt;img src="https://camo.githubusercontent.com/f2bdf5331e945f9e41e00ec42131818fc4f2ab27e8d0e320e4de2d493d0f5998/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f46726f6e74656e642d5265616374253230253242253230566974652d3230323332413f6c6f676f3d7265616374266c6f676f436f6c6f723d363144414642" alt="Frontend"&gt;&lt;/a&gt;
&lt;a rel="noopener noreferrer nofollow" href="https://camo.githubusercontent.com/018bde7b798190bb43564c53c0d6d417fe5f37ac54b7fd384d3cb0766e63013a/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4f6c6c616d612d4c6f63616c25323041492d6c69676874677265793f6c6f676f3d6f6c6c616d61266c6f676f436f6c6f723d626c61636b"&gt;&lt;img src="https://camo.githubusercontent.com/018bde7b798190bb43564c53c0d6d417fe5f37ac54b7fd384d3cb0766e63013a/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4f6c6c616d612d4c6f63616c25323041492d6c69676874677265793f6c6f676f3d6f6c6c616d61266c6f676f436f6c6f723d626c61636b" alt="Ollama"&gt;&lt;/a&gt;
&lt;a rel="noopener noreferrer nofollow" href="https://camo.githubusercontent.com/aa67ba18615aa719e8193a0bbf689c311df15539c3bec08f4c2c61af6c9de8f8/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4d6f64656c2d47656d6d612532303325323034422d3432383546343f6c6f676f3d676f6f676c65266c6f676f436f6c6f723d7768697465"&gt;&lt;img src="https://camo.githubusercontent.com/aa67ba18615aa719e8193a0bbf689c311df15539c3bec08f4c2c61af6c9de8f8/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4d6f64656c2d47656d6d612532303325323034422d3432383546343f6c6f676f3d676f6f676c65266c6f676f436f6c6f723d7768697465" alt="Gemma"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;a rel="noopener noreferrer" href="https://github.com/muskan-110/Sidewalk-Safari/docs/screenshots/main_page.png"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2Fmuskan-110%2FSidewalk-Safari%2FHEAD%2Fdocs%2Fscreenshots%2Fmain_page.png" alt="Landing screen"&gt;&lt;/a&gt;
&lt;a rel="noopener noreferrer" href="https://github.com/muskan-110/Sidewalk-Safari/docs/screenshots/activepage.png"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2Fmuskan-110%2FSidewalk-Safari%2FHEAD%2Fdocs%2Fscreenshots%2Factivepage.png" alt="Active safari"&gt;&lt;/a&gt;
&lt;a rel="noopener noreferrer" href="https://github.com/muskan-110/Sidewalk-Safari/docs/screenshots/complete.png"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2Fmuskan-110%2FSidewalk-Safari%2FHEAD%2Fdocs%2Fscreenshots%2Fcomplete.png" alt="Completion and field journal"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Features&lt;/h2&gt;
&lt;/div&gt;
&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Start an adventure:&lt;/strong&gt; choose 5, 10, or 15 minutes; campus, park/garden, neighborhood, or anywhere; easy, curious, or explorer difficulty.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;AI mission generator:&lt;/strong&gt; three short, safe missions (title, task, hint, estimated time, "what you might learn"), validated before they reach the screen.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Mission cards:&lt;/strong&gt; reveal-hint, mark-complete, a progress meter, a timer, and a field note after each find.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Pocket Mode:&lt;/strong&gt; a dark, minimal screen with only the timer…&lt;/li&gt;
&lt;/ul&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/muskan-110/Sidewalk-Safari" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;The README has one-minute setup instructions, the API, and the architecture.&lt;/p&gt;

&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;The open model:&lt;/strong&gt; Gemma 3 4B (&lt;code&gt;gemma3:4b&lt;/code&gt;), served locally through &lt;a href="https://ollama.com" rel="noopener noreferrer"&gt;Ollama&lt;/a&gt;. It's an open-weight model released under Google's Gemma Terms of Use (open weights, not an OSI open-source license). It's a 3.3 GB download, and I ran everything on an AMD Ryzen laptop with about 16 GB of RAM and no GPU.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Architecture:&lt;/strong&gt; React + Vite + Bootstrap 5 → FastAPI → Ollama, all on localhost. The journal is saved in the browser's localStorage. The model name is an environment variable (&lt;code&gt;TEXT_MODEL&lt;/code&gt;), and a single small function talks to Ollama, so the model underneath is a configuration choice.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;On a CPU, generation runs at about 10 tokens per second, so a set of missions takes around 45 seconds.&lt;/strong&gt; I treated the wait as part of the design: a "scouting your safari" screen with rotating walking tips, a seconds counter, and a nudge to put the phone in your pocket the moment the missions arrive.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What the backend does with every request:&lt;/strong&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Picks three focus themes in code, from groups, so you never get two leaf missions in one set.&lt;/li&gt;
&lt;li&gt;Builds a short prompt and asks Gemma for JSON.&lt;/li&gt;
&lt;li&gt;Validates the result with Pydantic: exactly three missions, length limits on every field.&lt;/li&gt;
&lt;li&gt;Runs a safety filter that rejects any mission telling you to touch, smell, pick, or close your eyes.&lt;/li&gt;
&lt;li&gt;Retries once, telling the model why it was rejected.&lt;/li&gt;
&lt;li&gt;Scales the estimated minutes so they never add up to more than your walk.&lt;/li&gt;
&lt;li&gt;If the model is unavailable or fails twice, serves a built-in mission pack and labels it clearly as not AI-generated.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;strong&gt;Three things I learned building with a small local model:&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Constrained decoding changed the writing.&lt;/strong&gt; With Ollama's JSON-schema &lt;code&gt;format&lt;/code&gt;, roughly one in three missions came out with words fused together ("youcan", "beennearby"). Asking for JSON in the prompt and validating with Pydantic instead produced none in an 18-mission test. I haven't pinned down exactly why, but it was easy to find because I could change how the model is called.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A prompt is not a safety system.&lt;/strong&gt; Gemma kept saying "close your eyes" on listening missions even when told not to. A code-level validator now enforces it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Variety comes from code, not from asking nicely.&lt;/strong&gt; A 4B model collapses into the same three ideas unless you hand it different themes.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The frontend uses an earthy, game-like look (chunky buttons, "FOUND!" stamps, leaf slots that fill as you go). Fonts and styles are bundled with the app, so nothing needs the internet at runtime.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;It runs on a laptop with no GPU, no API key and no per-request cost.&lt;/strong&gt; Mission generation happens on your computer, so there's no hosted service to be rate-limited, billed, or shut down.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Your data stays with you.&lt;/strong&gt; The browser, the FastAPI server and Ollama all run on localhost. The only thing the server ever receives is three choices: walk length, surroundings, and difficulty. Your journal never leaves your browser.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;I could change how the AI behaves, not just call it.&lt;/strong&gt; I rewrote the prompt, added my own validation and safety rules, read the model's real token speed from Ollama, and fixed the fused-word problem by changing how the model is called. With a closed API I'd have been guessing at all of that.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Model freedom.&lt;/strong&gt; The model is one environment variable, so the app isn't married to one provider's model, price, or terms.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;A hosted API would answer faster. For an app whose whole point is that you walk away from the screen, I'll take a 45-second wait in exchange for privacy, zero cost, and full control.&lt;/p&gt;

&lt;h2&gt;
  
  
  Prize Categories
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Best Use of Gemma:&lt;/strong&gt; Gemma 3 4B runs locally through Ollama and writes every mission. I measured its speed on CPU-only hardware, worked around its quirks with constrained-decoding testing and code-level safety validation, and built the app so it stays usable if the model fails.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Explain This Thing: An AI That Helps My Parents Understand Their Devices</title>
      <dc:creator>Muskan Khatoon</dc:creator>
      <pubDate>Sun, 04 Oct 2026 20:47:38 +0000</pubDate>
      <link>https://dev.to/muskan110/explain-this-thing-an-ai-that-helps-my-parents-understand-their-devices-3a35</link>
      <guid>https://dev.to/muskan110/explain-this-thing-an-ai-that-helps-my-parents-understand-their-devices-3a35</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-weekend-2026-10-01"&gt;Hacktoberfest Weekend Challenge: Build for a Friend&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

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

&lt;p&gt;My parents often use household devices like the AC, microwave, TV, and dishwasher, but they sometimes forget what particular buttons do.&lt;/p&gt;

&lt;p&gt;Usually, they would have to ask me:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;“What does this button do?”&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;And because these devices have so many buttons and modes, remembering them isn't always easy.&lt;/p&gt;

&lt;p&gt;So I built &lt;strong&gt;Explain This Thing&lt;/strong&gt; — a small personalized AI assistant that lets them upload a photo of a remote or control panel or used a saved photo, tap the button they're confused about, and get a simple explanation in English or Hindi.&lt;/p&gt;

&lt;p&gt;The goal wasn't to build an AI that understands every appliance in the world.&lt;/p&gt;

&lt;p&gt;I wanted to build something that understands the few devices my parents actually use.&lt;/p&gt;

&lt;p&gt;The app currently focuses on four devices:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;❄️ Bedroom AC Remote&lt;/li&gt;
&lt;li&gt;🍽️ Dishwasher&lt;/li&gt;
&lt;li&gt;🍲 Panasonic Microwave&lt;/li&gt;
&lt;li&gt;📺 Samsung TV Remote&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;They can select their device, use a saved photo or upload a new one, tap a button, and get an explanation of what it does and what they should press.&lt;/p&gt;

&lt;h3&gt;
  
  
  The part that mattered most
&lt;/h3&gt;

&lt;p&gt;I gave the app to my parents and asked them to try it themselves.&lt;/p&gt;

&lt;p&gt;Their feedback was simple:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;It helped them understand the buttons without needing my or anyone else's help.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;They also told me that they often forget what certain buttons do, so being able to quickly look it up themselves was useful.&lt;/p&gt;

&lt;p&gt;That was exactly the problem I wanted to solve.&lt;/p&gt;

&lt;p&gt;The project isn't about replacing a manual. It's about giving someone a quick way to regain that little bit of independence when they forget something.&lt;/p&gt;

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

&lt;p&gt;🎥 Demo video: &lt;a href="https://drive.google.com/file/d/1Fneh7OW8dZGmtfM7jQ_0ph0bTnEW-gEt/view?usp=sharing" rel="noopener noreferrer"&gt;https://drive.google.com/file/d/1Fneh7OW8dZGmtfM7jQ_0ph0bTnEW-gEt/view?usp=sharing&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The demo shows:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Selecting one of the saved devices&lt;/li&gt;
&lt;li&gt;Using a saved device photo&lt;/li&gt;
&lt;li&gt;Tapping a button&lt;/li&gt;
&lt;li&gt;Getting a simple explanation&lt;/li&gt;
&lt;li&gt;Switching between English and Hindi&lt;/li&gt;
&lt;li&gt;Using device-specific knowledge&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;

&lt;p&gt;GitHub: &lt;a href="https://github.com/muskan-110/Explain_This_Thing_HacktoberFest26" rel="noopener noreferrer"&gt;https://github.com/muskan-110/Explain_This_Thing_HacktoberFest26&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project is structured as a React frontend and FastAPI backend, with the device-specific knowledge stored separately for each appliance.&lt;/p&gt;

&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;The frontend is built with:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;React&lt;/li&gt;
&lt;li&gt;Vite&lt;/li&gt;
&lt;li&gt;TypeScript&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The backend uses:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Python&lt;/li&gt;
&lt;li&gt;FastAPI&lt;/li&gt;
&lt;li&gt;RapidOCR&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The AI runs locally using:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Ollama&lt;/li&gt;
&lt;li&gt;Gemma 3 4B for vision and explanation&lt;/li&gt;
&lt;li&gt;nomic-embed-text for embeddings&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  How a request works
&lt;/h3&gt;

&lt;p&gt;The basic pipeline looks like this:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;          Photo of device
                 │
                 ▼
        User taps a button
                 │
                 ▼
          Crop button area
                 │
                 ▼
              OCR
                 │
                 ▼
       Identify button label
                 │
                 ▼
       Search device knowledge
                 │
                 ▼
           Gemma 3 4B
                 │
                 ▼
       Simple explanation
                 │
          ┌──────┴──────┐
          ▼             ▼
       English        हिन्दी
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;If OCR can't read a label, for example on an icon-only button, Gemma's vision takes over and reads the label or icon instead.&lt;/p&gt;

&lt;p&gt;The app also keeps personalized knowledge for each device.&lt;/p&gt;

&lt;p&gt;For example:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;data/
└── appliances/
    ├── ac_remote/
    ├── dishwasher/
    ├── microwave/
    └── tv_remote/
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Each device has its own notes and information about its buttons.&lt;/p&gt;

&lt;p&gt;This means the AI doesn't have to rely entirely on general knowledge. It can retrieve information specifically written for that particular device.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;I deliberately built this around an open-weight model running locally rather than sending my parents' device photos to a closed AI API.&lt;/p&gt;

&lt;p&gt;The core AI runs through:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Ollama → Gemma 3 4B
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;This matters for this project because the input can contain photos of things inside someone's home.&lt;/p&gt;

&lt;p&gt;With local inference:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Photo
  ↓
My computer
  ↓
Ollama
  ↓
Gemma 3 4B
  ↓
Answer
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;instead of:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Photo
  ↓
External AI API
  ↓
Remote server
  ↓
Answer
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;That gives me more control over where the data is processed.&lt;/p&gt;

&lt;p&gt;It also means I don't need to pay for an API request every time my parents want to look up a button.&lt;/p&gt;

&lt;p&gt;Most importantly, using an open-weight model means I can experiment with the model, change the surrounding pipeline, replace the model, and build the experience around my users rather than being locked into one closed API.&lt;/p&gt;

&lt;p&gt;For this particular project, local AI isn't just a technical choice. It fits the problem.&lt;/p&gt;

&lt;h2&gt;
  
  
  Final note
&lt;/h2&gt;

&lt;p&gt;I started this project thinking about AI models and computer vision.&lt;/p&gt;

&lt;p&gt;I ended it thinking more about independence.&lt;/p&gt;

&lt;p&gt;Sometimes the useful thing you can build with AI isn't something spectacular.&lt;/p&gt;

&lt;p&gt;Sometimes it's just something that means your parents don't have to call you the next time they forget what a button does.&lt;/p&gt;

&lt;p&gt;And that's exactly why I built Explain This Thing. ❤️&lt;/p&gt;

&lt;h2&gt;
  
  
  Prize Categories
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Best Use of Gemma:&lt;/strong&gt; the core of the project is Gemma 3 4B running locally through Ollama. It reads icon-only buttons when OCR can't read a label, writes the explanations in simple English and Hindi, and returns structured output the app can display reliably.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>weekendchallenge</category>
      <category>hf26challenge</category>
    </item>
  </channel>
</rss>
