DEV Community: RUDRARAJSINH RANA

👻 The Last Login — A Private AI Digital Estate Planner

RUDRARAJSINH RANA — Sat, 31 Oct 2026 15:49:39 +0000

This is a submission for the "Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine" (https://dev.to/challenges/hacktoberfest-week4-2026-10-26)

👻 The Last Login — Your Memories. Your Wishes. Your Control.

What I Built

Imagine having thousands of photographs, personal documents, voice recordings, letters, and important memories stored across your devices.

But what happens to all these digital belongings when you are no longer able to manage them?

Who will know what to preserve, what to share, and what should remain private?

That question inspired me to build The Last Login, a privacy-first, local AI-powered digital estate planner.

The Last Login helps people organize their digital belongings, preserve the stories behind them, document personal wishes, and prepare an encrypted digital legacy archive for the future.

I interpreted the Ghost in the Machine theme as the digital traces we leave behind — the memories, stories, and personal information that continue to exist in our devices.

✨ Key Features

Life Inventory: Organize and manage important digital belongings.
Living Séance: Use local AI to help document the stories and memories behind personal items.
Keep, Release & Exorcise: Decide what to preserve, prepare for sharing, or exclude.
Legacy Letters: Create personal letters, stories, and instructions with AI assistance.
Sealed Ofrenda: Create an encrypted archive of selected legacy content and restore it later.
Memorial Preview: Preview only the content explicitly approved by the user.

Our philosophy is simple: AI assists, but humans decide.

We don't recreate or impersonate deceased people. We preserve real memories and personal wishes.

Demo

🚀 Live Application: [https://the-last-login-green.vercel.app/]

🎥 Demo Video: [PASTE YOUR DEMO VIDEO URL, IF AVAILABLE]

Code

💻 GitHub Repository: [https://github.com/rudrarajsinh920/The-Last-Login]

The project is built with a modular architecture using React, TypeScript, Node.js, Express, and SQLite.

How I Built It

I built The Last Login using modern open-source technologies, with a strong focus on local AI processing, privacy, and user control.

AI — Gemma through Ollama

I chose Gemma because I wanted to explore running AI locally instead of depending entirely on external cloud AI APIs.

Using Ollama, the application can run a compatible Gemma model on the user's machine.

I use local AI to assist with documenting memories and generating personal content.

When the local model is unavailable, supported operations can use an offline fallback.

Technology Stack

Frontend: React, TypeScript, Vite, Tailwind CSS
Backend: Node.js, Express.js, TypeScript
Database: SQLite with FTS5
AI: Gemma through Ollama
Encryption: AES-256-GCM and PBKDF2-SHA256
Validation: Zod

Security and Privacy

For the encrypted archive, I implemented AES-256-GCM encryption and password-based key derivation using PBKDF2-SHA256.

Users select what they want to include in their archive and remain responsible for protecting their passwords and recovery information.

The application is designed around local-first processing and explicit user decisions.

Why Does Open Innovation Matter?

Open innovation made it possible for me to experiment with local AI rather than building the entire experience around a closed cloud API.

Using Gemma and Ollama helped me explore how AI-powered applications can work closer to the user's own device.

It also gives developers more flexibility to experiment with models, customize AI workflows, and build applications without making every AI interaction dependent on an external paid API.

For a project dealing with personal memories and sensitive digital belongings, that flexibility matters.

I believe open innovation makes it possible to build technology that gives users more control over their data and how AI interacts with it.

My Agent Session

[OPTIONAL: ADD YOUR DEVRELAY AGENT SESSION LINK OR EMBED YOUR SESSION HERE]

The Spirit of It

The most meaningful part of building The Last Login was thinking about the stories behind ordinary digital files.

A photograph might look like just another image in a folder, but for someone, it could represent an important family memory.

A voice recording might contain a story that deserves to be preserved.

A personal letter might carry wishes that someone wants their loved ones to understand.

I wanted to build something that respects those memories instead of turning them into another dataset for AI.

The Last Login is not about bringing people back through AI. It is about helping people preserve what matters while they are here to make those choices themselves.

Your memories. Your wishes. Your control.

Prize Categories

[ADD THE ELIGIBLE PARTNER PRIZE CATEGORIES YOU ARE ACTUALLY ENTERING, BASED ON THE OFFICIAL CHALLENGE RULES.]

Built with curiosity, open-source AI, and the belief that our digital lives deserve thoughtful preservation.

devchallenge #hf26challenge

BillSloth

RUDRARAJSINH RANA — Sun, 25 Oct 2026 11:13:43 +0000

🦥 BillSloth — Your Calm, AI-Powered Bill Assistant

Less time managing paperwork. More time for what matters.

🛠️ What I Built

BillSloth is an AI-powered bill and invoice management assistant designed to make financial organization simpler, calmer, and more reliable.

Instead of searching through scattered documents or manually remembering every deadline, users can import bills, review extracted information, track due dates, and understand their outstanding bills in one organized workspace.

The idea behind BillSloth is simple: make bill management feel less like work.

✨ Key Features

📄 Bill & Invoice Import: Import supported PDF documents and images.
🔍 Information Extraction: Extract relevant bill details for review before saving.
📅 Due-Date Tracking: Keep track of upcoming and overdue bills.
📊 Financial Overview: Review outstanding bills and summaries based on stored records.
🤖 Ask BillSloth with Gemma: Use Gemma to support natural-language interactions and help users understand their bill records.
🔔 Reminders & Weekly Digest: Review upcoming deadlines and bill summaries.
🛡️ Audit Trail: Keep a record of supported actions for better traceability.
💾 Structured Storage: Store bill records using SQLite.

🚀 Demo

Live Demo: https://bill-sloth.vercel.app/

💻 Code

GitHub Repository: https://github.com/rudrarajsinh920/BillSloth

The source code is available for developers who want to explore the implementation, learn from it, or contribute improvements.

⚙️ How I Built It

I built BillSloth using a modern web development stack:

Frontend: React, TypeScript, Vite, and Tailwind CSS
Backend: Node.js and Express.js
Database: SQLite
Document Processing: PDF text extraction and OCR for supported images
AI Model: Google Gemma
AI Integration: Gemma integrated into BillSloth to support natural-language interactions about bill records

I designed the application around a practical workflow: import a bill, review the extracted information, save verified details, and use those records to manage deadlines and understand outstanding bills.

Gemma adds a natural-language AI layer to the experience. The application combines AI capabilities with structured records and application logic so that financial summaries can be grounded in stored bill data.

An important design decision was to keep essential financial calculations grounded in application data rather than relying on AI to invent totals.

💡 Why Does Open Innovation Matter?

Open innovation enables developers to learn from shared work, experiment with open technologies, and build tools that others can inspect, improve, and adapt.

By integrating Gemma into BillSloth, I explored how open AI models can support a practical everyday application rather than simply generating text or building another generic chatbot.

BillSloth demonstrates my interest in combining AI with traditional software engineering to create tools that are useful, understandable, and designed around real user needs.

By sharing the project openly, I hope other developers can explore its architecture, learn from its implementation, and contribute improvements.

🧠 Challenges & What I Learned

Building BillSloth helped me explore the connection between AI and traditional software engineering.

Some important lessons from the project were:

AI-generated information needs verification before it becomes a financial record.
Database queries and application logic are essential for reliable financial summaries.
Document extraction and natural-language understanding solve different parts of the problem.
AI integrations need error handling and clear feedback when a model is unavailable.
Privacy claims must accurately reflect where documents and AI requests are processed.

🤖 My Agent Session (Optional)

[Add your shareable Antigravity or AI coding-agent session link, if available.]

🦥 The Lazy Payoff

BillSloth is built around one simple idea: organizing bills should not feel like another full-time job.

Import a document, review its details, and keep your records organized in one place. Ask questions about your bills and spend less time searching through paperwork or trying to remember deadlines.

Slow and steady. Bills under control.

🏆 Prize Categories

Overall Challenge: Enter according to the official Week 3 rules and eligibility requirements.

Featured Technology / Partner Category: [Include a category only if the submitted project genuinely satisfies its official requirements.]

🔧 Run Locally

Visit the GitHub repository for the README, setup instructions, environment configuration, and local development commands.

hf26challenge

🌿 FIELDNOTE

RUDRARAJSINH RANA — Sun, 11 Oct 2026 07:54:51 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built
🌿 What is FIELDNOTE?
“The better you use FIELDNOTE, the less you need to look at your phone.”

FIELDNOTE is a local-AI observation engine designed for the Hacktoberfest 2026 Week 1 “Touch Grass” Challenge. Rather than competing for your screen time with streaks, push notifications, and gamification, FIELDNOTE is an intentional physical instrument that prompts you to put your phone away, step outside, and observe the living world around you.

Demo

fieldnote-97c7.onrender.com

Code

rudrarajsinh920 / fieldnote

🌿 FIELDNOTE

“AI that teaches you to notice.”
A collectible field journal from the future for Hacktoberfest 2026 Week 1 “Touch Grass” Challenge.

🔗 Live Application URL: https://fieldnote-97c7.onrender.com/

Core Product Philosophy

“The better you use FIELDNOTE, the less you need to look at your phone.”

FIELDNOTE is not another outdoor GPS tracker, gamified streak app, or generic conversational AI chatbot. It is a local-AI observation engine designed to minimize screen time and maximize physical sensory presence.

The Naturalist Ritual Loop

PREPARE (Define sensory window & intention)
   ↓
FIELD INVESTIGATION (Local AI generates an observation directive)
   ↓
FIELD MODE (7 seconds of inactivity → interface quietly disappears)
   ↓
PHONE AWAY (Observe physical reality with your senses)
   ↓
RETURN (Record raw notes, optional specimen photo & voice note)
   ↓
NATURALIST REFLECTION (Local AI distills ecological dynamics & inquiry)
   ↓
THE LOGBOOK (Permanently sealed into collectible on-device journal)

System Architecture

FIELDNOTE is architected…

View on GitHub

How I Built It
🔄 The Naturalist Ritual Loop
Prepare: Select a sensory lens (School & Science, Community & Environment, or Everyday Discovery) and define your outdoor observation window (e.g., 10 minutes).
Investigation Directive: Local open-weight AI (Ollama llama3.2:3b) synthesizes a grounded sensory prompt tailored to noticing patterns, textures, lichen symbiosis, leaf venation, or micro-habitats.
Field Mode (Screen Disappears): After 7 seconds of inactivity, the digital UI quietly dims and disappears, freeing your attention to explore nature with your actual senses.
Physical Observation: Explore the wilderness, park, or urban pavement with your phone in your pocket.
Return & Capture: Return to record sensory notes, capture macro specimen photography, or dictation recordings.
Naturalist Synthesis: Local AI distills observations into ecological dynamics, distinguishing verifiable evidence from hypothesis.
Collectible Logbook: Permanently sealed into an on-device naturalist folio with SQLite persistence and offline PWA capabilities.
🛠️ Technical Stack & Architecture
Frontend (@fieldnote/client): React 19, TypeScript, Vite, Tailwind CSS, Lucide icons, responsive from 375px mobile to 1440px+ desktop.
PWA & Offline Capability: Service Worker caching via Workbox for complete offline fieldwork in remote areas without signal.
Backend (@fieldnote/server): Node.js 24, Express, TypeScript, native node:sqlite database with Write-Ahead Logging (WAL).
Local AI Engine: Open-weight inference via Ollama (llama3.2:3b). 100% private, on-device computation with zero cloud tracking or telemetry.
Test Suite: Automated test suite with 10/10 passing tests validating AI recovery, timeout handling, state machine invariants, and SQLite storage.
Design System: Botanical editorial aesthetic inspired by vintage field guides, warm ivory paper tones (#FCFBF7), deep forest green (#12372A), and high-resolution landscape artwork.
Why Does Open Innovation Matter?

My Agent Session

Prize Categories

StudyMate Offline: A Privacy-First AI Study Companion That Works Without Internet 🎓

RUDRARAJSINH RANA — Sun, 04 Oct 2026 10:44:29 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

StudyMate Offline: A Privacy-First AI Study Companion That Works Without Internet 🎓

What I Built

I built StudyMate Offline, a privacy-first AI study companion, designed specifically for a college friend who has a daily two-hour commute with dead cell service and frequently studies in our campus library's basement where Wi-Fi drops out.

Like many students, they faced three major barriers with existing AI study tools:

Connectivity barriers: Cloud study tools break down the second you lose Wi-Fi on a train, flight, or in a basement study room.
Academic privacy risks: Discomfort with uploading private lecture materials, draft assignments, and thesis research to third-party commercial AI clouds where data can be logged or scraped.
Access costs: Expensive \$20–\$40/month subscriptions that students simply cannot afford.

StudyMate Offline addresses these challenges by bringing AI-powered study assistance directly to the student's personal computer.

It allows students to:

Upload and organize lecture notes (PDF and text files).
Search documents using semantic similarity powered by local vector embeddings.
Ask questions and receive explanations strictly grounded in their course notes.
Inspect verifiable source citations ([Source 1]) linked directly to exact text snippets.
Generate practice MCQs with instant grading and explanatory rationales.
Create personalized study plans calibrated to exam dates and available daily hours.

Its core mode runs 100% locally via Ollama using lightweight open-weight models (llama3.2:3b and nomic-embed-text). After downloading the models once, the entire study experience operates with airplane mode enabled—zero internet required.

The philosophy is simple: Your study materials should remain yours, and learning should never stop when the Wi-Fi does.

Demo

Live Cloud Preview: https://studymate-offline.onrender.com
Judges' 5-Minute Local Reproduction Guide: docs/JUDGES_REPRODUCTION_GUIDE.md

(Note: The hosted Render preview features an automatic cloud AI fallback so judges can test it online 24/7. The full, 100% offline, zero-data-transmission experience is demonstrated through the local setup guide).

Code

rudrarajsinh920 / studymate-offline

A privacy-first, locally runnable AI study companion for students (100% offline RAG with Ollama)

StudyMate Offline 🎓

A privacy-first, locally runnable AI study companion for students.
Grounded document Q&A with source citations, semantic vector search, practice quiz generation, and adaptive study planning powered entirely by local open-source models (Ollama).
100% offline. Zero cloud data transmission. Zero subscription fees.

📑 Table of Contents

🚨 Problem Statement

Modern students and researchers rely heavily on digital learning aids, but current AI study assistants suffer from severe structural flaws:

Academic Surveillance & Privacy Insecurity: Commercial cloud LLMs retain, inspect, and train on uploaded user documents. Students uploading proprietary…

View on GitHub

GitHub Repository: https://github.com/rudrarajsinh920/studymate-offline
License: MIT Open Source
Architecture: TypeScript monorepo with React frontend, Express backend, SQLite storage, and local Ollama inference.

How I Built It

StudyMate Offline is engineered for edge execution on consumer-grade laptops without discrete GPUs:

AI Models & Runtime

Meta's Llama 3.2 (3B): Run via Ollama for conceptual explanations, study planning, and quiz generation.
Nomic Embed Text: Generates 768-dimensional embeddings for high-precision semantic document retrieval.

Technology Stack

Frontend: React 18, TypeScript, Vite, Tailwind CSS, Lucide React
Backend: Node.js, Express, TypeScript, Multer, pdf-parse
Database: SQLite (WAL mode) for document chunks, binary vector BLOB caching, chat sessions, and quiz attempts
AI Runtime: Ollama (Local) with optional Gemini Flash fallback for 24/7 cloud hosting

Key Engineering Features

Grounded Retrieval-Augmented Generation (RAG): The system retrieves relevant document excerpts and strictly constrains model context before generating answers.
Source Transparency: Answers feature inline [Source N] citations that allow students to verify the exact source passage in their notes.
Honest Refusal Guardrails: When relevant evidence is missing from the uploaded notes, the tutor explicitly declines to answer (insufficientEvidence: true) instead of hallucinating false facts.
Context & Token Budgeting: Enforces concise response budgets (256–512 tokens) and top-$k$ filtering to keep CPU inference times under ~30 seconds on standard laptops.
100% Test Coverage: Verified with 18/18 milestone integration tests and 10/10 automated system audit workflows passing.

Why Does Open Innovation Matter?

Open innovation made the core concept of StudyMate Offline possible:

Privacy & Data Ownership: Students process personal academic materials locally instead of surrendering their data to commercial AI providers.
Accessibility & Equity: Open-weight models democratize AI education, eliminating the need for expensive monthly subscriptions.
Uninterrupted Learning: Learning shouldn't depend on network stability or cloud server uptime.
Freedom to Customize: Open model ecosystems allow developers to tailor inference and prompts directly to student needs and laptop hardware.
Trust Through Transparency: Verifiable citations prevent students from blindly trusting generic AI outputs during high-stakes exam prep.

Open innovation is not just about free software—it enables developers to build technology around the real-world constraints of everyday students.

Prize Categories

Best Overall "Build for a Friend" Project
Best Use of Open-Source / Open-Weight AI