DEV Community

Cover image for CampusGemma AI: Turning Student Freelance Chaos into Profitable SOWs and Paid Mentorship with Google Gemma 2
Vimal Sahani
Vimal Sahani

Posted on

CampusGemma AI: Turning Student Freelance Chaos into Profitable SOWs and Paid Mentorship with Google Gemma 2

#devchallenge #hf26challenge #gemma #python

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

As a final-year IT student at IIIT Bhopal, I watch classmates and fellow junior developers try to launch their first side hustle every single semester. Some want to build freelance React landing pages for local businesses, others want to offer 1-on-1 mock interview preparation for juniors, and others do freelance Python scripting.

Almost all of them hit the exact same brick wall in their first week:

  1. The WhatsApp Brief Trap: Clients send chaotic, 15-sentence stream-of-consciousness WhatsApp messages with vague demands.
  2. The Unpaid Scope-Creep Spiral: Because students don't know how to write binding legal Statements of Work (SOW), clients request "just one more tiny change" indefinitely.
  3. The Proprietary API Tax: When students try adding AI to their side-hustle offerings, closed API token meters (OpenAI, Anthropic) eat away their micro-margins and instill billing anxiety.

To solve this, I built CampusGemma AI — a zero-marginal-cost micro-business studio powered by Google's open-weight Gemma 2 (gemma2-9b-it) and enterprise OpenTelemetry GenAI observability.

💼 The 4 Core Business Engines

  1. Freelance SOW & Scope Creep Shield: Turns messy client briefs into professional Statements of Work with itemized deliverables, milestone USD quotes, and explicit anti-scope-creep boundaries ("Any revision exceeding 2 iterations billed at $25/hr").
  2. Paid Mock-Interview Studio: Empowers senior students to monetize their placement experience by auto-generating structured technical interview rubrics, candidate rubrics, and answer keys.
  3. Client Code & Security Auditor: A pre-delivery safety net that audits client code for exposed secrets, security leaks, and maintainability before handing over deliverables.
  4. OpenTelemetry & Unit Economics Tracker: Real-time GenAI tracing that monitors latency, token usage, and confirms a 100% gross profit margin ($0.00 marginal inference cost).

Demo

CampusGemma AI Banner

Key Walkthrough:

  • Generate & Export SOW in 1 Click: Pick one of the quick presets (e.g., Local Bakery Online Ordering or Student Club QR Ticketing), adjust your hourly rate, and click Generate Formal SOW & Quote. Within seconds, Gemma 2 produces a bulletproof client agreement with milestones and scope boundaries, ready to download via the "📥 Download Formal SOW Agreement (.md)" button.
  • Monetized Mentorship Scorecards: Run 1-on-1 technical placement interviews or security code audits and instantly export structured candidate scorecards.
  • Inspect Live OpenTelemetry Tracing: Open the Observability & Economics tab to inspect active OpenTelemetry spans, latency breakdowns, and profit margin metrics confirming 100% gross margins.

Code

The complete codebase is open-source under the Apache 2.0 license:

# Clone and run locally in under 60 seconds
git clone https://github.com/VimalN2005/campus-gemma-ai.git
cd campus-gemma-ai
pip install -r requirements.txt
streamlit run app.py
Enter fullscreen mode Exit fullscreen mode

Core Architecture Snippet (core/engine.py): 

def generate_freelance_proposal(client_brief: str, hourly_rate: int = 35) -> str:
    """Uses Google Gemma 2 to transform messy client chats into binding SOWs."""
    system_prompt = (
        "You are an elite freelance business advisor. Transform unstructured client briefs "
        "into structured, professional Statements of Work (SOW). Include: "
        "1. Executive Summary, 2. Itemized Deliverables, 3. Milestones & Pricing, "
        "and 4. Scope-Creep Protection clause."
    )
    user_prompt = f"Client Brief:\n{client_brief}\n\nTarget Rate: ${hourly_rate}/hr"
    return query_gemma_model(system_prompt, user_prompt)
Enter fullscreen mode Exit fullscreen mode

How I Built It

1. The Open-Weight Core: Google Gemma 2 (gemma2-9b-it)

I selected Google's Gemma 2 9B Instruct model. Gemma 2 offers an extraordinary balance of reasoning capability, structured JSON adherence, and lightweight footprint. It follows complex contract formatting instructions with precision that rivals models 5x its size.

2. High-Speed Inference Engine

The application queries Gemma 2 via Groq's high-speed LPU infrastructure for near-instant (<1.2s) generation, with a built-in deterministic fallback engine for offline campus hackathons.

3. OpenTelemetry GenAI Instrumentation

As an active contributor to the CNCF OpenTelemetry Python GenAI ecosystem, I wired custom span tracing (core/telemetry.py) into every inference call. Every business action records:

  • gen_ai.system: gemma2
  • gen_ai.request.model: gemma2-9b-it
  • Latency and token throughput
  • Calculated marginal cost ($0.00)

Why Does Open Innovation Matter?

Why build this on open-weight models rather than a closed API like GPT-4 or Claude?

  1. Zero Marginal Token Cost = Viable Micro-Hustles:

    When a student charges $60 to build a student club website, every cent matters. If they rely on closed proprietary APIs for iterative client drafts, testing, and debugging, unexpected token bills can wipe out their weekend earnings. With open-weight models like Gemma 2, inference is free on local GPUs or zero-tier cloud endpoints. The business model works at unit volume 1.

  2. Absolute Client Confidentiality & IP Security:

    Freelancers frequently handle sensitive client materials — unreleased wireframes, confidential business logic, and proprietary codebases. Closed APIs retain rights to train or inspect payloads unless you pay for expensive enterprise tiers. Gemma 2 can be run 100% locally on a developer laptop with Ollama, ensuring zero client data ever touches external servers.

  3. Freedom from Vendor Lock-In:

    If a commercial AI API changes its terms of service, raises prices, or bans a use case, a small side hustle dies overnight. With Gemma 2, the weights belong to the open-source community forever.

My First User

For this challenge, I didn't want to stop at code. I tested CampusGemma AI with a real person: Aman, a 3rd-year Computer Science student at IIIT Bhopal who takes on freelance frontend projects.

The Real Problem:

Aman was dealing with a local client who kept sending rambling WhatsApp voice notes asking for "a clean portal, maybe add an admin dashboard too, and let's make it look like Airbnb". Aman had originally verbally agreed to build a simple 2-page website for \$100, but the client was now demanding a database, authentication, and multiple redesigns — without extra compensation.

The Test:

We pasted the client's messy WhatsApp conversation into CampusGemma AI's Freelance SOW Generator.

Within two seconds, Gemma 2 generated:

  • A crystal-clear 3-milestone breakdown.
  • An adjusted quote of \$160 reflecting the extra database work.
  • An explicit boundary clause: "Out of Scope: Additional design revisions beyond 2 rounds billed at \$25/hr."

The Result:

Aman exported the SOW as PDF and sent it back to the client. The client replied within 30 minutes: "This looks very professional, let's stick to Milestone 1 & 2 for now, here is the 50% deposit."

Aman earned his first protected freelance dollar with zero unpaid scope creep!

Prize Categories

I am officially entering CampusGemma AI into the following categories:

  • Featured Category: Best Use of Gemma (#gemma) CampusGemma AI is built directly around Google's open-weight Gemma 2 model (gemma2-9b-it), demonstrating how open-weight AI enables student freelancers and indie developers to operate profitable, privacy-first side hustles with zero API marginal costs.
  • Overall Challenge Winner

Built with ❤️ during Hacktoberfest 2026. Feedback and contributions are warmly welcome on GitHub!

Top comments (0)