DEV Community: Akash

LedgerLens: A Privacy-First AI Expense Auditor for Freelancers | Built with Open-Source AI & Render

Akash — Fri, 16 Oct 2026 19:55:31 +0000

What I Built

I built LedgerLens, an open-source, privacy-first AI expense auditor for freelancers. It automatically categorizes invoices and receipts, detects duplicate charges and calculation errors, flags potential tax compliance issues, estimates tax savings, and generates financial reports. Built with FastAPI, Gemma 2, and Ollama, it supports local processing to help protect sensitive financial data.

Demo

ledgerlens-ph9u.onrender.com

Code

Akash10053218 / LedgerLens

🔍 LedgerLens

Privacy-First AI Invoice & Expense Auditor for Freelancers and Solopreneurs.
Built with open-source AI and open-weight models at its core. Zero client data leaks. 100% offline-ready or deployable directly to Render.

💡 The Problem: The Freelancer's Invoice Dilemma

Freelancers, developers, and agency owners juggle dozens of receipts and client invoices each month. Traditional accounting tools (QuickBooks, Expensify) present major friction:

Confidentiality & NDA Violations: Closed-source cloud accounting platforms upload itemized client invoices (showing client names, secret rates, project code names, and bank details) to third-party proprietary AI platforms.
Expensive SaaS Lock-In: Charging $35–$70/month for simple document sorting and basic exports.
Flawed Tax Classification: Generic OCR tools fail to apply specific IRS Schedule C rules (e.g., the 50% business meal deduction rule under IRC § 274(n), Form 1099-NEC contractor filings, and the $75 receipt archiving requirement).

⚡ The Solution: LedgerLens

LedgerLens is a dedicated side-hustle…

View on GitHub

How I Built It

I built LedgerLens using FastAPI, Python, and JavaScript, with PyPDF and Pillow for document processing. I integrated Gemma 2 and Ollama for local AI-powered expense analysis, alongside rule-based logic for tax classification, duplicate detection, and compliance checks. The frontend provides an interactive dashboard, tax savings simulator, and CSV/Markdown exports, while Render Blueprint simplifies deployment.

Why Does Open Innovation Matter?

Open innovation makes AI tools accessible, affordable, and transparent. LedgerLens allows developers to improve and customize the software while helping freelancers manage finances without expensive subscriptions or unnecessary reliance on closed AI services. It encourages community collaboration, data privacy, and accessible technology for everyone.

My Agent Session

I used an AI-assisted development workflow to plan LedgerLens, structure its backend, integrate open-source AI models, and refine expense auditing features. The focus was on building a practical, privacy-first tool with reliable tax rules, automated checks, and a simple user experience.

My First User

My first user was myself. I built LedgerLens to solve the invoice and expense-tracking problems I faced as a freelancer. Testing it on my own workflow helped me focus on practical features like automatic categorization, error detection, and simple financial reports.

Prize Categories

Best Use of Render: Uses Render Blueprint (render.yaml) to simplify deployment of the Ledger

GamePath — Your Game, Your Path

Akash — Sun, 04 Oct 2026 18:03:22 +0000

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

What I Built

GamePath is a clean progression tracker and roadmap viewer for gamers who want one place to track their PlayStation game completions.
I built it for a friend who loves 100% completing games but hates switching between wiki pages, YouTube walkthroughs, and different guides.

GamePath brings everything together in one dashboard with game metadata, YouTube walkthroughs and boss fights, AI-generated roadmaps, missions, and collectible checklists. Players can mark their progress as they play, with progress saved in their browser's local storage.

Demo

gamepath.onrender.com

Code

Akash10053218 / GamePath

A Project which helps gamers get the roadmap and tutorials of the games they want to play.

GamePath

Game tutorials, roadmaps, and walkthroughs for PlayStation and Xbox games.

GamePath reads a live PS/Xbox catalog from RAWG, provides click-to-load YouTube searches for video references, and displays hand-curated or pre-generated AI walkthroughs. Progress is stored only in the current browser with localStorage.

This is an open-source MVP. There are no accounts, database, or request-time AI generation.

Features

RAWG catalog limited to PS4, PS5, Xbox One, and Xbox Series
Search, platform and genre filters, sorting, and Load more pagination
Game details: cover, genre, release year, playtime, developer, publisher, description, and screenshots
Chapter, mission, boss, collectible, and tip walkthrough data
data/games.json hand-curated roadmaps override generated walkthrough files
Click-to-load YouTube search embeds for full walkthroughs, boss fights, and collectibles
Mission completion, collectible tracking, and progress bars saved in localStorage
Responsive dark interface

Tech stack

Next.js App Router + TypeScript
Tailwind CSS
RAWG API for game metadata
Gemini, Groq, and OpenRouter free…

View on GitHub

How I Built It

GamePath is built with Next.js, React, and Tailwind CSS, with a responsive gamer-focused interface.

Game data comes from the RAWG API, while AI is used to generate structured walkthroughs, missions, tips, and collectible checklists.
The AI generation pipeline uses models and providers such as Llama 3.1 8B, Groq, OpenRouter, and Gemini 1.5 Flash. The generated content is structured as JSON and validated using Zod.

Why Does Open Innovation Matter?

Open innovation and open-weight AI models make projects like GamePath much more accessible.

Models such as Llama 3 allow developers to generate and process large amounts of structured content without relying entirely on expensive closed AI APIs. This makes it possible for independent developers and students to build useful, data-rich applications at a much lower cost.

My Agent Session

This project was built collaboratively with an AI coding agent, which helped with planning, implementation, debugging, and refining the GamePath MVP.

Prize Categories

I am entering GamePath into the following categories:
Best Use of Render — GamePath is configured for deployment on Render using a render.yaml Blueprint.
Best Use of Entire — GamePath was built collaboratively with an AI agent.