DEV Community

Rachit Singh
Rachit Singh

Posted on

ProposalPilot — Say Goodbye to Generic Upwork Bids

#devchallenge #hf26challenge

Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Submission 💸

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

I built ProposalPilot — an AI-powered proposal generator engineered specifically for freelancers, agency owners, and side-hustlers bidding on platforms like Upwork, Fiverr, Toptal, and direct cold outreach.

The Problem It Solves

Freelancers spend anywhere from 30 to 60 minutes writing a single custom bid. With average platform win rates hovering between 5% and 15%, freelancers end up spending 10–15 hours every week doing unpaid proposal writing. By the time they reach their 8th proposal of the day, fatigue sets in and proposals become generic, which tanks conversions even further.

Who Would Pay For It

Any active freelancer or digital agency bidding on client projects. At a modest $9/month subscription (or free self-hosted), ProposalPilot pays for itself with just one extra won contract per year. It cuts proposal crafting time down from 45 minutes to under 60 seconds while keeping pitch quality high and personalized.

Demo

  • Live Local Preview: Run locally in 60 seconds (npm run dev at http://localhost:3000)
  • Key Workflow:
    1. Paste Job Post: Paste raw job post requirements or RFP text from Upwork/Fiverr.
    2. Select Platform & Tone: Optimize for Upwork's direct hook, Fiverr's concise value pitch, or direct consultative proposals.
    3. Local or Open AI Generation: Instant local generation or high-depth Gemma inference.
    4. One-Click Export: Copy directly to clipboard or download formatted Markdown.

Code

The complete source code is open source under the MIT License:

GitHub logo rachiittt / ProposalPilot

🚀 ProposalPilot — AI Proposal Generator for Freelancers

Your side hustle wingman. Generate winning freelance proposals in seconds using open-source AI. Privacy-first, $0 running cost, 100% local.

ProposalPilot License Hacktoberfest

🎯 What It Does & Who Would Pay For It

ProposalPilot is an AI-powered proposal generator built specifically for freelancers on platforms like Upwork, Fiverr, Toptal, and direct client outreach.

The problem: Freelancers spend 30-60 minutes writing each proposal. With a typical 5-15% win rate, that's hours of unpaid work every week. Most proposals sound generic because after the 10th one today, you're out of creative energy.

The solution: Paste a job description → get a personalized, platform-optimized proposal in under 60 seconds. Your client data never leaves your machine, and it costs $0 to run.

Who would pay: Any freelancer bidding on projects. At $9/mo for the hosted version (vs. $0 self-hosted), it pays for itself if it wins…

View on GitHub

Repository link: https://github.com/rachiittt/ProposalPilot

Architecture Overview 

ProposalPilot/
├── index.html          # Modern glassmorphic web app UI
├── styles/main.css     # Dark mode design system & micro-animations
├── scripts/main.js     # Client-side NLP & local proposal generator
├── server/index.js     # Lightweight backend for Gemma open-weight LLM inference
├── vite.config.js      # Build tooling and local dev server
└── README.md           # Documentation & setup guides
Enter fullscreen mode Exit fullscreen mode

How I Built It

ProposalPilot is built around open-weight AI models and a local-first architecture:

  1. Open-Weight AI Core (Gemma):

    • Built around Google's open-weight model family (Gemma 3 27B IT via @google/genai).
    • Gemma analyzes the project scope, identifies unstated client pain points, extracts key technical deliverables, and crafts a persuasive bid with custom portfolio talking points.

  2. Offline-First Fallback Engine:

    • The client application features a built-in heuristic NLP engine in pure JavaScript that extracts keywords, budget tiers, and timelines offline without external API dependencies.
    • When offline or running without an API key, users still get high-conversion structured proposals at zero latency and zero cost.

  3. Frontend & Design System:

    • Built with Vite, semantic HTML5, and vanilla modern CSS.
    • Features a custom dark-mode interface with glassmorphism, responsive controls, live proposal character counters, and real-time streaming text effects.

Why Does Open Innovation Matter?

Open innovation is essential for a tool like ProposalPilot for three main reasons:

1. Privacy is Non-Negotiable

Freelancers constantly handle confidential project specs, proprietary client requirements, NDA-governed details, and custom pricing strategies. Passing client bids through black-box, closed APIs means sending sensitive commercial data to third-party servers where it may be logged or used for model training. With open-weight models running locally or through private instances, client data never leaves the developer's machine.

2. Viable Unit Economics ($0 Marginal Cost)

Closed AI APIs charge per input/output token. Bidding on 20–30 jobs a day quickly accumulates $15–$45/month in API token costs alone. For a side hustler just starting out, that eats up profits before landing their first client. Open-weight models like Gemma democratize AI access: they have $0 per-generation cost and can be run self-hosted or on local hardware.

3. Freedom From Vendor Lock-In

Closed APIs can alter their terms of service, deprecate endpoints, or increase token pricing overnight. Building on open-weight models allows developers to fine-tune weights on their own winning proposal history, swap runtimes, and maintain complete ownership of their workflow.

Top comments (0)