<?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: Kunal Badgujar</title>
    <description>The latest articles on DEV Community by Kunal Badgujar (@kunal_badgujar_07).</description>
    <link>https://dev.to/kunal_badgujar_07</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%2F4157111%2Fc810bda0-2fc2-4f9c-a1cf-5a34e350e2ae.png</url>
      <title>DEV Community: Kunal Badgujar</title>
      <link>https://dev.to/kunal_badgujar_07</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/kunal_badgujar_07"/>
    <language>en</language>
    <item>
      <title>Locentra</title>
      <dc:creator>Kunal Badgujar</dc:creator>
      <pubDate>Sat, 31 Oct 2026 15:00:10 +0000</pubDate>
      <link>https://dev.to/kunal_badgujar_07/locentra-3nlg</link>
      <guid>https://dev.to/kunal_badgujar_07/locentra-3nlg</guid>
      <description>&lt;h2&gt;
  
  
  Locentra: Making the Invisible World of Indoor Positioning Visible 👻📡
&lt;/h2&gt;

&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week4-2026-10-26"&gt;Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;Locentra&lt;/strong&gt; is an interactive indoor-localisation research and simulation platform designed to explore how wireless positioning systems estimate the location of people and objects inside buildings.&lt;/p&gt;

&lt;p&gt;Inspired by the theme &lt;em&gt;Ghost in the Machine&lt;/em&gt;, Locentra explores the invisible signals and mathematical intelligence operating behind indoor positioning.&lt;/p&gt;

&lt;p&gt;GPS works well outdoors, but indoor environments introduce challenges such as signal reflections, obstacles, packet loss, and inaccurate distance measurements.&lt;/p&gt;

&lt;p&gt;Locentra provides a virtual laboratory where these challenges can be simulated, visualised, and evaluated.&lt;/p&gt;

&lt;h3&gt;
  
  
  Key Features
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Interactive 2D &amp;amp; 3D Spatial Lab:&lt;/strong&gt; Visualise indoor rooms, positioning anchors, and target trajectories.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Wireless Measurement Simulation:&lt;/strong&gt; Simulate Gaussian noise, packet dropouts, NLOS bias, and measurement outliers.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multiple Positioning Algorithms:&lt;/strong&gt; Compare Non-Linear Least Squares, Robust Huber Estimation, and Kalman Filtering.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Benchmark Arena:&lt;/strong&gt; Evaluate algorithms using RMSE, MAE, P95 error, and valid-estimate rates.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Scenario Studio:&lt;/strong&gt; Configure room geometry, anchor arrangements, and simulation parameters.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Dataset Explorer:&lt;/strong&gt; Explore and analyse generated positioning datasets.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The idea is to make invisible wireless measurements visible, understandable, and measurable.&lt;/p&gt;

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

&lt;p&gt;🚀 &lt;strong&gt;Live Application:&lt;/strong&gt; &lt;a href="https://locentra-olive.vercel.app/" rel="noopener noreferrer"&gt;https://locentra-olive.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;⚙️ &lt;strong&gt;Live Backend API:&lt;/strong&gt; &lt;a href="https://locentra.onrender.com/" rel="noopener noreferrer"&gt;https://locentra.onrender.com/&lt;/a&gt;&lt;/p&gt;

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

&lt;p&gt;💻 &lt;strong&gt;GitHub Repository:&lt;/strong&gt; &lt;a href="https://github.com/badgujarkunal93-blip/Locentra" rel="noopener noreferrer"&gt;https://github.com/badgujarkunal93-blip/Locentra&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project is open source, and contributions, suggestions, and improvements are welcome!&lt;/p&gt;

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

&lt;p&gt;Locentra combines a modern web interface with a Python-based numerical positioning engine.&lt;/p&gt;

&lt;h3&gt;
  
  
  Frontend
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;React 19&lt;/li&gt;
&lt;li&gt;TypeScript&lt;/li&gt;
&lt;li&gt;Vite&lt;/li&gt;
&lt;li&gt;Tailwind CSS&lt;/li&gt;
&lt;li&gt;Three.js&lt;/li&gt;
&lt;li&gt;React Three Fiber&lt;/li&gt;
&lt;li&gt;React Three Drei&lt;/li&gt;
&lt;li&gt;Zustand&lt;/li&gt;
&lt;li&gt;TanStack React Query&lt;/li&gt;
&lt;li&gt;Recharts&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Backend
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Python&lt;/li&gt;
&lt;li&gt;FastAPI&lt;/li&gt;
&lt;li&gt;NumPy&lt;/li&gt;
&lt;li&gt;SciPy&lt;/li&gt;
&lt;li&gt;Pandas&lt;/li&gt;
&lt;li&gt;Pydantic&lt;/li&gt;
&lt;li&gt;Uvicorn&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Deployment
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Vercel:&lt;/strong&gt; Frontend hosting and deployment.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Render:&lt;/strong&gt; Backend hosting and FastAPI deployment.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  How the System Works
&lt;/h3&gt;

&lt;p&gt;Virtual Room Configuration → Anchor Placement → Synthetic Measurement Generation → Position Estimation → Algorithm Comparison → Performance Visualisation&lt;/p&gt;

&lt;p&gt;The numerical engine uses mathematical optimisation and filtering techniques to estimate target positions from simulated wireless distance measurements.&lt;/p&gt;

&lt;p&gt;Robust estimation techniques help reduce the impact of noisy and corrupted measurements, while the interactive interface makes algorithm behaviour easier to understand.&lt;/p&gt;

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

&lt;p&gt;Open innovation matters because advanced positioning technology should not remain limited to expensive hardware laboratories and proprietary systems.&lt;/p&gt;

&lt;p&gt;By building Locentra using open-source technologies such as NumPy, SciPy, FastAPI, React, and Three.js, the project makes experimentation and algorithm evaluation more accessible.&lt;/p&gt;

&lt;p&gt;Developers and researchers can inspect the implementation, experiment with different positioning algorithms, modify simulation parameters, and extend the platform.&lt;/p&gt;

&lt;p&gt;It also gives the community an opportunity to explore indoor localisation before investing in physical positioning hardware.&lt;/p&gt;

&lt;p&gt;The goal is to make research more transparent, reproducible, and accessible.&lt;/p&gt;

&lt;h2&gt;
  
  
  My Agent Session
&lt;/h2&gt;

&lt;p&gt;Locentra was developed through an AI-assisted workflow involving architecture planning, implementation, debugging, testing, and iterative refinement.&lt;/p&gt;

&lt;p&gt;The development process involved validating the numerical positioning engine, improving the simulation pipeline, and refining the interactive visualisation experience.&lt;/p&gt;

&lt;h2&gt;
  
  
  The Spirit of It
&lt;/h2&gt;

&lt;p&gt;Imagine entering a building where GPS stops working.&lt;/p&gt;

&lt;p&gt;Your device still exists. The room still exists. But locating your exact position becomes much harder.&lt;/p&gt;

&lt;p&gt;Behind the scenes, invisible wireless signals travel between anchors and devices. Locentra recreates that invisible world through mathematics, simulations, and interactive visualisation.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The ghost is not the device. It is the invisible intelligence working behind the signals.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Locentra is my attempt to make that invisible intelligence visible.&lt;/p&gt;

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

&lt;h3&gt;
  
  
  Overall Challenge
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Hacktoberfest Open-Source AI Challenge — Week 4: Ghost in the Machine&lt;/strong&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Partner Categories
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Render ($200)&lt;/strong&gt; — Locentra uses Render to deploy and host its Python FastAPI backend. The service exposes health-check and simulation endpoints, connecting the frontend to the numerical positioning engine through HTTP APIs.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of GitHub Copilot ($100)&lt;/strong&gt; — AI-assisted development of the Locentra codebase, including implementation, debugging, and refinement. &lt;em&gt;(Include this category only if GitHub Copilot was actually used.)&lt;/em&gt;
&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>SARAS AI: Stop Studying Twice — Learn Once, Recall Better</title>
      <dc:creator>Kunal Badgujar</dc:creator>
      <pubDate>Sat, 24 Oct 2026 13:11:55 +0000</pubDate>
      <link>https://dev.to/kunal_badgujar_07/saras-ai-stop-studying-twice-learn-once-recall-better-4hgf</link>
      <guid>https://dev.to/kunal_badgujar_07/saras-ai-stop-studying-twice-learn-once-recall-better-4hgf</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week3-2026-10-19"&gt;Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h1&gt;
  
  
  SARAS AI — Learn. Recall. Understand.
&lt;/h1&gt;

&lt;p&gt;What if finishing a lecture didn't mean spending another hour preparing to revise it?&lt;/p&gt;

&lt;p&gt;Students often watch an educational video, then switch between tools to write notes, create flashcards, draw diagrams, and prepare questions. The actual learning can get buried under the work required to organize it.&lt;/p&gt;

&lt;p&gt;I built &lt;strong&gt;SARAS AI&lt;/strong&gt; to tackle that problem.&lt;/p&gt;

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

&lt;p&gt;SARAS AI is an AI-powered learning workspace designed to transform passive video watching into structured, active learning. It brings learning resources and revision tools together so students can spend less time preparing to study and more time understanding what they learn.&lt;/p&gt;

&lt;h3&gt;
  
  
  Key features
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Smart Video Ingestion:&lt;/strong&gt; Organize educational YouTube lectures into structured learning chapters.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Checkpoint Quiz Gating:&lt;/strong&gt; Test your understanding before progressing through a lesson.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Smart Notes:&lt;/strong&gt; Turn lesson content into organized notes, definitions, examples, and summaries.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Flashcards and Active Recall:&lt;/strong&gt; Practice remembering concepts instead of repeatedly rereading them.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Mind Maps and Flowcharts:&lt;/strong&gt; Visualize relationships between ideas and break down complex processes.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Interactive Knowledge Graph:&lt;/strong&gt; Explore connections between topics and their prerequisites.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Socratic AI Tutor:&lt;/strong&gt; Get guidance that supports your learning process.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Exam Planner:&lt;/strong&gt; Organize revision around upcoming exams and available study time.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The central idea is simple: &lt;strong&gt;automate repetitive preparation without automating away the learning itself.&lt;/strong&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Try the Demo
&lt;/h2&gt;

&lt;p&gt;🚀 &lt;strong&gt;Live demo:&lt;/strong&gt; &lt;a href="https://sarasai.vercel.app" rel="noopener noreferrer"&gt;https://sarasai.vercel.app&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Explore the interface and see how SARAS AI brings lesson content, quizzes, notes, and revision tools into one learning workspace.&lt;/p&gt;

&lt;h2&gt;
  
  
  Explore the Code
&lt;/h2&gt;

&lt;p&gt;💻 &lt;strong&gt;GitHub repository:&lt;/strong&gt; &lt;a href="https://github.com/badgujarkunal93-blip/sarasai" rel="noopener noreferrer"&gt;https://github.com/badgujarkunal93-blip/sarasai&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;SARAS AI is an open-source project. The repository contains the source code, technical documentation, setup instructions, and AI integration architecture.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Tech stack:&lt;/strong&gt; Next.js, TypeScript, Tailwind CSS, Mermaid.js, Prisma ORM, and a modular AI integration layer.&lt;/p&gt;

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

&lt;p&gt;I approached SARAS AI as a learning workflow rather than another general-purpose chatbot.&lt;/p&gt;

&lt;p&gt;The application is organized around a sequence: ingest educational content, structure it into chapters, support understanding with learning resources, and help students revise through active recall.&lt;/p&gt;

&lt;p&gt;The AI integration layer is designed to support tasks such as:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Transcript segmentation and chapter organization&lt;/li&gt;
&lt;li&gt;Chapter-based quiz generation&lt;/li&gt;
&lt;li&gt;Smart notes and flashcards&lt;/li&gt;
&lt;li&gt;Mind map and flowchart generation&lt;/li&gt;
&lt;li&gt;Image-based learning assistance&lt;/li&gt;
&lt;li&gt;Concept extraction and tutor conversations&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;For visual learning resources, Mermaid.js turns supported diagram definitions into rendered diagrams. The architecture also provides a path for local inference through a compatible model runtime, while a deterministic fallback helps the application demonstrate selected workflows when inference is unavailable.&lt;/p&gt;

&lt;p&gt;That separation makes it easier to experiment with different model configurations without redesigning the entire learning experience.&lt;/p&gt;

&lt;p&gt;One important distinction: local inference requires a compatible model, runtime, and configuration. The hosted demo should not be mistaken for a fully offline application, and AI-dependent features can vary depending on the active inference configuration.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Open Innovation Matters
&lt;/h2&gt;

&lt;p&gt;I believe open innovation matters because developers should have room to understand, adapt, and improve the technology behind the products they build.&lt;/p&gt;

&lt;p&gt;For SARAS AI, that flexibility has three important benefits.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;1. The freedom to experiment&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;A modular AI architecture makes it easier to evaluate different open-weight models, experiment with prompts, and adapt structured outputs to educational tasks.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. A path toward local-first learning&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Supporting compatible local inference creates a foundation for exploring learning tools that can run on a student's own machine. With the appropriate model and hardware, this could reduce dependence on hosted inference and offer more control over where processing happens.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;3. Building beyond a black box&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Open models and adaptable tooling let developers investigate how their systems work, modify their workflows, and build specialized applications around real needs.&lt;/p&gt;

&lt;p&gt;For me, the goal is not to add AI just because it is trending. It is to build something useful with an architecture that can evolve as models and tools improve.&lt;/p&gt;

&lt;h2&gt;
  
  
  The Lazy Payoff
&lt;/h2&gt;

&lt;p&gt;The biggest time-waster isn't always studying. Sometimes, it's preparing to study.&lt;/p&gt;

&lt;p&gt;After a programming lecture, a student might spend additional time writing summaries, creating revision questions, drawing process diagrams, and deciding what to review first.&lt;/p&gt;

&lt;p&gt;SARAS AI aims to bring those activities into one workspace, giving students a structured starting point for revision instead of making them start from scratch every time.&lt;/p&gt;

&lt;p&gt;That doesn't mean removing the effort required to learn. Students still need to solve problems, test their understanding, and apply concepts independently.&lt;/p&gt;

&lt;p&gt;I call this &lt;strong&gt;productive laziness&lt;/strong&gt;: letting technology handle repetitive preparation so people can focus on the work that actually builds understanding.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Less busywork. More active learning.&lt;/strong&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Want to Improve Next
&lt;/h2&gt;

&lt;p&gt;SARAS AI is a work in progress, and there is room to make it more useful.&lt;/p&gt;

&lt;p&gt;My next priorities are to improve the reliability of AI-generated learning resources, make model configuration clearer, strengthen the local-inference path, and continue refining the learning workflow based on real student feedback.&lt;/p&gt;

&lt;p&gt;I also want to make it easier for contributors to understand the architecture and help extend the project.&lt;/p&gt;

&lt;h2&gt;
  
  
  Final Thoughts
&lt;/h2&gt;

&lt;p&gt;Building SARAS AI has reinforced an idea I care about: useful AI should do more than generate text. It should help people complete meaningful workflows.&lt;/p&gt;

&lt;p&gt;For students, that means spending less time organizing learning materials and more time understanding, remembering, and applying what they learn.&lt;/p&gt;

&lt;p&gt;For developers, it means keeping the system adaptable enough to experiment, improve, and build on open technology.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;SARAS AI — Learn. Recall. Understand.&lt;/strong&gt;&lt;/p&gt;

&lt;h1&gt;
  
  
  OpenSourceAI #Hacktoberfest #AI #EdTech
&lt;/h1&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>ai</category>
      <category>opensource</category>
    </item>
    <item>
      <title>TrailMate AI</title>
      <dc:creator>Kunal Badgujar</dc:creator>
      <pubDate>Mon, 12 Oct 2026 12:55:38 +0000</pubDate>
      <link>https://dev.to/kunal_badgujar_07/trailmate-ai-4nii</link>
      <guid>https://dev.to/kunal_badgujar_07/trailmate-ai-4nii</guid>
      <description>&lt;h1&gt;
  
  
  TrailMate AI: Your Offline Nature Companion 🌿
&lt;/h1&gt;

&lt;p&gt;&lt;em&gt;An open-source AI companion that helps people explore nature, identify birds, and spend less time on screens.&lt;/em&gt;&lt;/p&gt;

&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;TrailMate AI is a concept for an AI-powered outdoor companion designed to make nature exploration easier and more engaging.&lt;/p&gt;

&lt;p&gt;The idea is simple: instead of spending hours on a phone, people can use AI to discover birds, learn about plants, plan outdoor walks, and document their experiences in nature.&lt;/p&gt;

&lt;p&gt;Key features include:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Bird Identification:&lt;/strong&gt; Identify birds from photographs using an open-source vision model.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Nature Explorer:&lt;/strong&gt; Learn about plants, trees, and wildlife around you.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Offline-First Experience:&lt;/strong&gt; Run supported AI models locally so basic features can work without an internet connection.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Trail Journal:&lt;/strong&gt; Record observations and keep a personal log of outdoor discoveries.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Screen-Time Friendly:&lt;/strong&gt; Encourage users to explore outdoors rather than remain glued to their devices.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  🎬 Demo
&lt;/h2&gt;

&lt;p&gt;This submission describes the proposed concept and workflow. A working demonstration and live deployment are not yet available.&lt;/p&gt;

&lt;h2&gt;
  
  
  💻 Code
&lt;/h2&gt;

&lt;p&gt;The project is intended to use open-source components, with a lightweight interface and locally runnable AI models.&lt;/p&gt;

&lt;p&gt;The proposed implementation would include a responsive web interface, an open-source vision model for wildlife recognition, and local storage for nature observations.&lt;/p&gt;

&lt;h2&gt;
  
  
  🛠️ How I Would Build It
&lt;/h2&gt;

&lt;p&gt;My proposed technology stack:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Frontend:&lt;/strong&gt; React and Tailwind CSS&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;AI:&lt;/strong&gt; An open-weight vision-language model for image understanding&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Local Inference:&lt;/strong&gt; Ollama or another compatible local inference runtime&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Storage:&lt;/strong&gt; Browser-based local storage&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Deployment:&lt;/strong&gt; Vercel for the web interface&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The goal is to keep the architecture modular so users can replace models and adapt the application to their own devices.&lt;/p&gt;

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

&lt;p&gt;Nature exploration should not depend on a paid API, a constant internet connection, or sending personal photographs to a third-party server.&lt;/p&gt;

&lt;p&gt;An open-source approach could let users run supported models locally, inspect how the system works, change the model, and contribute improvements. Local processing can also help protect privacy by keeping observations on the user's device.&lt;/p&gt;

&lt;p&gt;Open innovation makes AI more accessible and gives developers the freedom to experiment without depending entirely on a single closed platform.&lt;/p&gt;

&lt;h2&gt;
  
  
  🤖 My Agent Session
&lt;/h2&gt;

&lt;p&gt;No agent session is attached to this concept submission.&lt;/p&gt;

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

&lt;p&gt;Primary theme: &lt;strong&gt;Touch Grass&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The concept focuses on open-source AI, local inference, privacy, and encouraging real-world outdoor exploration.&lt;/p&gt;

&lt;h2&gt;
  
  
  🌍 Final Thoughts
&lt;/h2&gt;

&lt;p&gt;AI should not always demand more of our attention. Sometimes, it should help us put our phones away and appreciate the world around us.&lt;/p&gt;

&lt;p&gt;TrailMate AI is a concept built around that idea: use open technology to make outdoor discovery more accessible, private, and fun.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Less scrolling. More exploring. More nature. 🌿&lt;/strong&gt;&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>StudyMate — Your AI Study Partner for Exam Prep</title>
      <dc:creator>Kunal Badgujar</dc:creator>
      <pubDate>Sun, 04 Oct 2026 13:29:49 +0000</pubDate>
      <link>https://dev.to/kunal_badgujar_07/studymate-your-ai-study-partner-for-exam-prep-3kb5</link>
      <guid>https://dev.to/kunal_badgujar_07/studymate-your-ai-study-partner-for-exam-prep-3kb5</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;I built &lt;strong&gt;StudyMate&lt;/strong&gt;, an AI study partner designed for a friend who was preparing for engineering finals and struggling with dense technical subjects.&lt;/p&gt;

&lt;p&gt;The problem wasn't a lack of study material.&lt;/p&gt;

&lt;p&gt;It was knowing &lt;strong&gt;what to study, what to practice, and what they actually didn't understand&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;StudyMate turns their study material into a complete learning loop:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Notes → Study Guide → AI Quiz → Performance → Weak Topics → Targeted Revision&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Instead of simply asking an AI chatbot questions, the student can give StudyMate their own notes and let the system build an exam-focused study session around them.&lt;/p&gt;

&lt;h3&gt;
  
  
  The experience
&lt;/h3&gt;

&lt;ol&gt;
&lt;li&gt;Add study material.&lt;/li&gt;
&lt;li&gt;Generate an exam-oriented study guide with Gemma.&lt;/li&gt;
&lt;li&gt;Take an AI-generated quiz based only on that material.&lt;/li&gt;
&lt;li&gt;Get deterministic scoring.&lt;/li&gt;
&lt;li&gt;See which topics need more work.&lt;/li&gt;
&lt;li&gt;Generate targeted revision for those weak topics.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The goal was simple:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;Don't just help my friend study more. Help them study the things they actually need to improve.&lt;/strong&gt;&lt;/p&gt;
&lt;/blockquote&gt;

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

&lt;p&gt;🚀 &lt;strong&gt;Live Demo:&lt;/strong&gt; &lt;a href="https://studymate-sable-zeta.vercel.app/" rel="noopener noreferrer"&gt;https://studymate-sable-zeta.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;If the hosted AI backend is not available in the deployed demo environment, here's the recorded demo showing the complete working flow:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;[Add your demo video link here]&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The demo flow is:&lt;/p&gt;

&lt;p&gt;&lt;code&gt;Add Material → Generate Study Guide → Generate Quiz → Answer Questions → Identify Weak Topics → Targeted Revision&lt;/code&gt;&lt;/p&gt;

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

&lt;p&gt;💻 &lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/badgujarkunal93-blip/StudyMate" rel="noopener noreferrer"&gt;https://github.com/badgujarkunal93-blip/StudyMate&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The repository contains the complete frontend, backend, Gemma integration, local storage layer, planning documents, and setup instructions.&lt;/p&gt;

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

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

&lt;ul&gt;
&lt;li&gt;React 19&lt;/li&gt;
&lt;li&gt;TypeScript&lt;/li&gt;
&lt;li&gt;Vite&lt;/li&gt;
&lt;li&gt;Node.js&lt;/li&gt;
&lt;li&gt;Express&lt;/li&gt;
&lt;li&gt;IndexedDB&lt;/li&gt;
&lt;li&gt;localStorage fallback&lt;/li&gt;
&lt;li&gt;Google's &lt;code&gt;@google/genai&lt;/code&gt; SDK&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Gemma 4 — &lt;code&gt;gemma-4-26b-a4b-it&lt;/code&gt;&lt;/strong&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The frontend provides the study experience while the Node/Express backend keeps the AI API credentials away from the browser.&lt;/p&gt;

&lt;p&gt;The architecture is:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;                    STUDYMATE
                        │
              ┌─────────┴─────────┐
              │                   │
         Study Library        AI Actions
              │                   │
        IndexedDB             Node/Express
              │                   │
              │              @google/genai
              │                   │
              │                Gemma 4
              │                   │
              └─────────┬─────────┘
                        ↓
                Study Guide / Quiz
                        ↓
                 User Performance
                        ↓
                  Weak Topics
                        ↓
                Targeted Revision
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  Why Gemma?
&lt;/h3&gt;

&lt;p&gt;Gemma is the core AI model behind StudyMate.&lt;/p&gt;

&lt;p&gt;I use it for:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Exam-oriented study guide generation&lt;/li&gt;
&lt;li&gt;Key concept extraction&lt;/li&gt;
&lt;li&gt;Important terminology&lt;/li&gt;
&lt;li&gt;AI-generated practice questions&lt;/li&gt;
&lt;li&gt;Topic classification&lt;/li&gt;
&lt;li&gt;Targeted revision&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The quiz itself does &lt;strong&gt;not&lt;/strong&gt; ask an AI model to decide whether the student was correct. Gemma generates the questions and their answer indices, while StudyMate performs the actual scoring deterministically in the browser.&lt;/p&gt;

&lt;p&gt;That separation makes the learning loop more predictable.&lt;/p&gt;

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

&lt;p&gt;This was one of the most important decisions in the project.&lt;/p&gt;

&lt;p&gt;I didn't want to build another application where the AI layer was just an invisible call to a proprietary chatbot.&lt;/p&gt;

&lt;p&gt;The challenge gave me a reason to explore what an &lt;strong&gt;open-weight model&lt;/strong&gt; could make possible.&lt;/p&gt;

&lt;p&gt;StudyMate uses &lt;strong&gt;Gemma&lt;/strong&gt;, an open-weight model, as the intelligence behind the core learning workflow.&lt;/p&gt;

&lt;p&gt;That matters because the AI isn't just an optional chatbot feature.&lt;/p&gt;

&lt;p&gt;It defines how the application works:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Material → Gemma → Learning resources → Quiz → Weakness analysis → Revision&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Using an open-weight model also gives the project a much more flexible foundation. The model can be experimented with, swapped, adapted, and integrated into different environments instead of treating the intelligence layer as an untouchable black box.&lt;/p&gt;

&lt;p&gt;I also had a practical constraint: my laptop isn't powerful enough to comfortably run a large model locally.&lt;/p&gt;

&lt;p&gt;So instead of pretending that local inference was practical, I used hosted Gemma while keeping the student's study library in the browser.&lt;/p&gt;

&lt;p&gt;The privacy model is intentionally explicit:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;Your study library stays on your device. Material is sent to Gemma only when you request AI assistance.&lt;/strong&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;That was a better fit for the real person I was building for.&lt;/p&gt;

&lt;h2&gt;
  
  
  What Makes StudyMate Different?
&lt;/h2&gt;

&lt;p&gt;A lot of AI study tools stop here:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Upload notes → Get summary&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;StudyMate goes further:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;YOUR NOTES
    ↓
UNDERSTAND
    ↓
PRACTICE
    ↓
MEASURE
    ↓
FIND WEAKNESS
    ↓
FIX IT
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The application doesn't just generate content.&lt;/p&gt;

&lt;p&gt;It creates a &lt;strong&gt;feedback loop&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;If a student repeatedly performs poorly on a topic, that topic becomes part of their weak-topic radar. They can then ask StudyMate to generate focused revision material specifically for that area.&lt;/p&gt;

&lt;p&gt;That's the part I wanted my friend to actually use before an exam.&lt;/p&gt;

&lt;h2&gt;
  
  
  Design
&lt;/h2&gt;

&lt;p&gt;I also wanted StudyMate to feel different from the typical "AI SaaS dashboard."&lt;/p&gt;

&lt;p&gt;So I built a &lt;strong&gt;retro student app × brutalist UI × modern AI product&lt;/strong&gt; aesthetic:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Dark charcoal background&lt;/li&gt;
&lt;li&gt;Chunky typography&lt;/li&gt;
&lt;li&gt;Red accents&lt;/li&gt;
&lt;li&gt;Thick borders&lt;/li&gt;
&lt;li&gt;Hard offset shadows&lt;/li&gt;
&lt;li&gt;Tactile buttons&lt;/li&gt;
&lt;li&gt;Responsive mobile navigation&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The interface is intentionally bold because studying doesn't need to look like another corporate dashboard.&lt;/p&gt;

&lt;h2&gt;
  
  
  Building It
&lt;/h2&gt;

&lt;p&gt;I built StudyMate as a completely new project for this challenge.&lt;/p&gt;

&lt;p&gt;I used an AI coding workflow with Antigravity and GSD Core to plan and implement the application in phases:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Phase 1:&lt;/strong&gt; Design system and application shell&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Phase 2:&lt;/strong&gt; Study material management&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Phase 3:&lt;/strong&gt; Real Gemma integration&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Phase 4:&lt;/strong&gt; AI quiz + weak-topic engine&lt;/p&gt;

&lt;p&gt;The GSD workflow helped me keep the project focused instead of turning a weekend challenge into an unnecessarily huge application.&lt;/p&gt;

&lt;p&gt;The final application has:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Persistent local study materials&lt;/li&gt;
&lt;li&gt;Real Gemma API integration&lt;/li&gt;
&lt;li&gt;Structured AI responses&lt;/li&gt;
&lt;li&gt;Server-side API key protection&lt;/li&gt;
&lt;li&gt;Deterministic quiz scoring&lt;/li&gt;
&lt;li&gt;Persistent quiz performance&lt;/li&gt;
&lt;li&gt;Weak-topic detection&lt;/li&gt;
&lt;li&gt;Targeted AI revision&lt;/li&gt;
&lt;li&gt;Responsive mobile/desktop UI&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  A Small Detail I Care About
&lt;/h2&gt;

&lt;p&gt;The application doesn't pretend that everything is AI.&lt;/p&gt;

&lt;p&gt;Gemma handles the parts where generative intelligence is useful.&lt;/p&gt;

&lt;p&gt;The application handles deterministic logic itself.&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;Gemma:&lt;/strong&gt;&lt;br&gt;&lt;br&gt;
"What questions should this student practice?"&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;StudyMate:&lt;/strong&gt;&lt;br&gt;&lt;br&gt;
"Did they actually answer question 4 correctly?"&lt;/p&gt;

&lt;p&gt;That separation makes the system more reliable and easier to reason about.&lt;/p&gt;

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

&lt;p&gt;I'm entering:&lt;/p&gt;

&lt;h3&gt;
  
  
  🏆 Best Use of Gemma
&lt;/h3&gt;

&lt;p&gt;StudyMate uses Google's &lt;strong&gt;Gemma 4 (&lt;code&gt;gemma-4-26b-a4b-it&lt;/code&gt;)&lt;/strong&gt; as the core AI model for study-guide generation, quiz generation, topic classification, and targeted revision.&lt;/p&gt;

&lt;p&gt;I am only claiming this category because Gemma is genuinely central to the application's functionality.&lt;/p&gt;

&lt;h2&gt;
  
  
  Final Thoughts
&lt;/h2&gt;

&lt;p&gt;I started with a very simple problem:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;My friend has plenty of notes. They don't have a good way to turn those notes into effective exam preparation.&lt;/strong&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;That became StudyMate.&lt;/p&gt;

&lt;p&gt;It isn't trying to replace a teacher.&lt;/p&gt;

&lt;p&gt;It isn't trying to be another general-purpose chatbot.&lt;/p&gt;

&lt;p&gt;It's a small AI study partner designed around one person's actual workflow:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Understand → Practice → Measure → Improve.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;And that's what made this challenge fun for me.&lt;/p&gt;

&lt;p&gt;Thanks for checking out StudyMate! 🚀&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Live Demo:&lt;/strong&gt; &lt;a href="https://studymate-sable-zeta.vercel.app/" rel="noopener noreferrer"&gt;https://studymate-sable-zeta.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Source:&lt;/strong&gt; &lt;a href="https://github.com/badgujarkunal93-blip/StudyMate" rel="noopener noreferrer"&gt;https://github.com/badgujarkunal93-blip/StudyMate&lt;/a&gt;&lt;/p&gt;

&lt;h1&gt;
  
  
  Hacktoberfest #AI #Gemma #React #TypeScript #DEVChallenge
&lt;/h1&gt;

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