DEV Community: krithikgokuls

GemmaPay: Privacy-First AI Invoicing & Client Follow-Up Studio Powered by Google Gemma

krithikgokuls — Tue, 13 Oct 2026 15:20:25 +0000

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

What I Built

GemmaPay is a privacy-first, open-weight AI side-hustle studio built specifically for freelancers, solopreneurs, and small service businesses. It automates two of the most time-consuming, uncomfortable, and costly operational tasks in freelancing:
Automated Client Invoicing: Instantly converts rough, messy work notes, hour logs, or milestone chats into legally structured, professional invoices with automated tax, discounts, deposits, and payment QR codes.
Personalized Service Follow-Ups & Retainer Engine: Generates context-aware, relationship-preserving communication sequences across 7 critical client stages (from gentle pre-due reminders and firm overdue escalations to post-project review requests and recurring retainer pitches).
Who Would Pay For It & What Problem Does It Solve?
Target Audience: Freelance developers, graphic designers, digital marketers, consultants, wedding photographers, and local service providers juggling client work alongside a 9-to-5 or building a full-time independent practice.
The Problem:
Over 60% of freelancers report chasing late payments, wasting an average of 4–6 hours every month writing awkward follow-up emails or drafting manual invoices.
Existing commercial SaaS tools (FreshBooks, HoneyBook, QuickBooks Self-Employed) charge recurring subscription fees (
39/month or
468/year)—eating directly into thin side-hustle margins.
Closed cloud software exposes sensitive client contracts, tax IDs, and confidential billing rates to proprietary third-party servers.
The Solution: GemmaPay delivers studio-grade billing, dynamic QR code payments, client health tracking, and multi-channel follow-up automation at $0 monthly software overhead, all while keeping sensitive client data private through open-weight AI.

Demo

Live Application:https://ais-pre-wqw35rwcvdjmbf3uaw7uqf-239371853037.asia-southeast1.run.app
Development Preview: https://ais-dev-wqw35rwcvdjmbf3uaw7uqf-239371853037.asia-southeast1.run.app

Core Interactive Features:
Smart Note Parser: Type or paste unstructured notes like "Design 5-page Shopify store for Coastal Brew, 15 hrs @ $65/hr, plus $120 plugin license, Net 15, 30% deposit already paid", and Gemma structures it into a verified invoice in seconds.
Printable & PDF-Ready Layouts: 4 professional design themes (Modern Emerald, Executive Navy, Minimal Mono, Studio Amber) with custom print-to-PDF formatting and generated payment QR codes (Stripe, PayPal, Venmo, ACH Wire).
Follow-Up Studio: Multi-tone generation with 1-click mailto: email launcher, quick SMS/WhatsApp snippets, and tactical upsell advice to convert one-off gigs into predictable monthly retainers.
Income & Rate Growth Dashboard: Live monthly income goal tracker ($/mo), effective hourly yield calculator, client lifetime value (CLV) matrix, and an AI package recommender.
Live Client PII Scrubber: Real-time privacy bench that flags and redacts credit cards, SSNs, and personal bank accounts.

Code

krithikgokuls / gemmapay

Hacktoberfest Dev Challenge 2

gemmapay

Privacy-first freelance invoice automation and personalized client follow-up studio powered by Google Gemma open-weight models with zero data retention.

View on GitHub

How I Built It

Open-Source AI & Models:
Google Gemma 2 (gemma-2-27b-it / gemma-2-9b-it): Google’s open-weight model family provides the core reasoning engine. Gemma extracts structured JSON entities from natural-language notes and generates polite, persuasive, and relationship-sensitive client correspondence.
Local Inference & Ollama Support: GemmaPay includes an architecture configuration for local, air-gapped execution via Ollama (ollama run gemma2:2b or ollama run gemma2:9b on localhost:11434), allowing freelancers to generate invoices and messages 100% offline with zero cloud dependency.
Heuristic Deterministic Fallback Engine: If offline or facing temporary network interruptions, an on-device rule engine ensures freelancers are never locked out of billing their clients.
Tech Stack:
Runtime & Backend: Node.js, Express, TypeScript, @google/genai SDK.
Frontend: React 19, Tailwind CSS, Lucide Icons, qrcode library.
Storage & Security: Browser localStorage for private persistence—no remote databases storing your bank accounts or client contacts.

Why Does Open Innovation Matter?

Open innovation and open-weight models like Google Gemma change the economics of running a small business:
Zero Data Retention for Financial Privacy:
Proprietary, closed APIs often log, inspect, or retain user prompts to train future commercial models. For freelancers handling NDA-bound client projects, hourly rates, and banking wire instructions, data exposure is a major liability. Open-weight models guarantee ephemeral execution and complete data sovereignty.
Eliminating the "SaaS Tax":
Closed subscription tools charge $20 to
250 to $450+ every year**.
True Portability & Anti-Fragility:
With open weights, the software does not depend on a single vendor's pricing tier, terms of service change, or API shutdown. If internet access drops on a train or in a cafe, Gemma can run on-device via Ollama without losing functionality.

My Agent Session

Development Session: Built, tested, and deployed end-to-end via Google AI Studio Build.
Model Engine: Google Gemma 2 / Gemini Developer Platform.
Key Iterations Handled:
Configured full-stack server architecture with Express mounting Vite middlewares.
Implemented print-specific CSS rules for pixel-perfect PDF invoice export.
Integrated dynamic payment QR code generation for Stripe, PayPal, Venmo, and bank wires.
Built multi-stage follow-up logic covering every phase from pre-due reminders to retainer upselling.

My First User

User: Marcus Vance, a freelance motion designer and video editor balancing evening client work with a day job.
Their Feedback:
"Chasing late invoices was the worst part of my week—I hated sounding aggressive, but I also needed to get paid on time. GemmaPay generated a friendly reminder in 5 seconds that got my $1,450 packaging invoice paid that afternoon. Plus, using the retainer pitch stage, I convinced that same client to book a $750/month ongoing video editing retainer. That's real income I didn't have before."
Outcome: Successfully collected a 4-day overdue balance of $1,000 and converted the relationship into a recurring monthly contract.

Prize Categories

Google Gemma Challenge / Open-Weight AI Category: Demonstrates practical utility, cost reduction, and privacy preservation using Google's Gemma model architecture.
Solo Creator / Freelancer Productivity Category: Directly empowers side-hustlers and small businesses to generate predictable revenue, eliminate SaaS subscription costs, and save hours of administrative overhead.

Team Submissions:

Creator / Lead Developer: @krithikgokuls (Solo Submission)

Thanks for participating!

Gemma TrailMate 🌿 — The Voice-First Outdoor AI Companion That Gets You Off the Screen

krithikgokuls — Wed, 07 Oct 2026 14:37:25 +0000

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

What I Built

Gemma TrailMate 🌿 is a voice-first outdoor AI companion built around one absolute principle: “The screen should be the shortest part of the experience.”
Existing AI assistants often pull us deeper into our screens. Gemma TrailMate does the opposite: it uses open-source AI intelligence to synthesize highly customized, short, and active outdoor missions, then collapses its entire interface into a minimalist countdown timer and checkbox sheet. It encourages users to lock their phones, slip them into their pockets, step out, experience nature, and return only when finished to record their experiences.

Demo

https://ais-pre-kdh7wmdrjmfra363y6dee5-718654813775.asia-southeast1.run.app

Code

https://github.com/krithikgokuls/gemma-trailmate

How I Built It

AI Layer: Google's Gemma open-weight model family. We created custom system prompts that mold the LLM specifically to Gemma's 9B IT persona and concise, active-voice tone.
Frontend: React, TypeScript, and Tailwind CSS v4 utilizing an earthy, minimal visual design.
Backend: Express.js proxy routes mounting Vite dev middleware to allow unified full-stack serving on a single port.
Offline Resilience: Offline caching logic automatically loads preset local missions and nature guides during trail connectivity drops.
Privacy Controls: Zero external tracking databases are used. Private journals and photos remain strictly inside the user's browser local storage.

Why Does Open Innovation Matter?

Getting people outside should not require sending every personal reflection, observation, or location coordinate to a closed, multi-billion-parameter cloud monopoly.
Open-weight models like Gemma put developers and users back in control:
Airgapped Privacy: Gemma's compact size allows it to run entirely locally (via Ollama or NPUs) in your browser or phone, guaranteeing your personal thoughts never leave your physical device.
Trailhead Portability: It enables applications to continue running fully offline in wilderness areas with zero network connectivity by hosting the weights on local trail routers.
Customizability: Developers can fine-tune Gemma specifically on regional botanical datasets, creating localized forest, desert, or mountain guides without being bound to generic closed-API rules.

My Agent Session

You can view the full agent development session and iterative build history here:
Google AI Studio Developer Session: cb86a0bf-3ab4-470b-93d9-24825a466628

Prize Categories

Grand Prize Category: Touch Grass AI / Open Source Innovation (Primary)
Gemma / Open-Weight AI Developer Track

StudyBuddy 🦉 – The Privacy-First, Swappable Local AI Study Companion I Built for My Friend

krithikgokuls — Sat, 03 Oct 2026 02:02:19 +0000

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

What I Built

I built StudyBuddy – AI Companion for My Friend, a responsive, offline-first study dashboard designed specifically for my friend Arun. Arun is preparing for dense technical certifications (specifically the Cisco CCNA & Computer Networks engineering exams) and frequently struggles with three common study bottlenecks:
Passive, high-stress reading of dry, technical documentation (such as RFCs, port mappings, and OSI layers).
The lack of interactive, safe, and personalized practice quizzes that diagnose why an answer is wrong, rather than just grading it.
Managing consistent study patterns and timelines leading up to high-stakes exam dates.
StudyBuddy solves this by converting static lecture slides and notes into an interactive, friendly educational sandbox. It features an AI Tutor that explains complex topics using real-world analogies (e.g., comparing TCP vs. UDP to registered letters vs. postcards), a Quiz Generator, a 3D Flashcard Reader utilizing spaced repetition triggers, an AI Study Scheduler, and a Knowledge Gap Diagnostics Analyzer that maps weak areas and compiles practice questions.

Demo

ais-dev-lo3flqbas6wdx5rovj3m4m-718654813775.asia-southeast1.run.app

Code

krithikgokuls / studybuddy-local-ai

🦉 Swappable Local AI study companion powered by Ollama & Llama 3.2. 100% private, offline-first exam preparation, quizzes, 3D flashcards, and tutor chat.

StudyBuddy – AI Companion for My Friend

StudyBuddy is a polished, highly responsive, privacy-first AI study companion built specifically to help students and friends prepare for exams and certifications (like the Cisco CCNA) with less stress and absolute confidence.

Designed around open-weight/open-source AI models (such as Llama 3.2), the application maintains a strict Offline-First / Local AI ethos. Study materials never leave your device unless you explicitly opt to route them through custom local servers.

🚀 Key Features

Dashboard: Comprehensive overview of study streaks, daily target goals, mastered modules, and weak diagnostics.
Materials Hub: Upload or paste text/Markdown notes locally with secure local storage containment.
AI Tutor: Real-time conversational tutoring Grounded strictly in your notes, utilizing simple real-world analogies to break down dry documentation. Includes macro triggers: Explain Simpler, Give me an Example, and Test me.
Quiz Generator: Synthesize multi-choice knowledge check questions…

View on GitHub

How I Built It

StudyBuddy is designed to break the vendor lock-in of expensive, closed cloud APIs.

Swappable AI Abstraction & Local Inference The default architecture is optimized to support Llama 3.2 (3B) or Mistral (7B) running locally on the user's laptop through Ollama (http://localhost:11434). We query Ollama's /api/chat and /api/generate endpoints. To handle structured tasks (like constructing quiz structures with incorrect option explanations or creating study plans with duration values), we pass structured JSON schemas and toggle Ollama's format: "json" parameter. Llama 3.2 interprets and parses these rules directly. Robust CORS Handling: Because browsers block cross-origin requests to local services by default, the app detects connection errors and provides explicit setup steps (e.g., launching with OLLAMA_ORIGINS="*" ollama serve) to make local-hosted inference seamless.
UI & Aesthetic Engineering The interface reflects custom, crafted design elements from the Education & Interactive Simulations guidelines: 60-30-10 Color Scheme: Soft, clean off-white canvas (#F8FAFC) representing 60% of background space; warm whites and orange borders (#FAF6F0) making up 30%; and bright terracotta accent buttons (#D96A43) drawing focus to action points. Zero-Pill Metadata: Follows strict typographical discipline, rendering date indicators and categories as clean text with typographic separators (·), avoiding colored capsule chips. Tactile Spaced Repetition: Flashcards incorporate pure CSS perspective transforms (rotateY(180deg)) for a clean, natural 3D flip animation when clicked. Cozy Graphics: Features custom-engineered visual graphics including an academic owl tutor mascot badge (studybuddy_mascot) and a cozy morning study desk header (dashboard_study_header).

Why Does Open Innovation Matter?

Open-source and open-weight AI is revolutionary for education, student privacy, and accessibility:
Absolute Privacy: School essays, draft notes, or code guidelines are deeply personal and proprietary. closed-source cloud APIs require uploading this information to remote servers, exposing private student details. Running Llama 3.2 locally via Ollama ensures Arun's materials never leave his physical hard drive.
True Offline Independence: Arun studies in library basements, on airplanes, and in transit where network connectivity is spotty or non-existent. A closed API fails instantly in these scenarios. Open-weight models running on local silicon turn Arun's personal computer into an offline, high-capacity tutor.
Infinite Customization & Zero Cost: Traditional cloud APIs charge per-token, which rapidly becomes unsustainable for students running thousands of practice quizzes. Open-weight inference is entirely free to run on student hardware, allowing limitless active recall loops.

My Agent Session

I built this entire application in an interactive coding session with the Google AI Studio Agent.

The agent helped me:

Model & Abstraction: Design the swappable AIProvider contract separating the local Ollama module from the offline sandbox.
Tactile UX Design: Scaffold the 3D perspective CSS code for the rotating flashcards and align the colors to a cozy educational theme.
Compilation & Quality: Iterate and compile the React code flawlessly, verifying it with rigorous, zero-error lint checks.

Prize Categories

Built for a Friend ❤️: Highly customized parameters specifically tuned to help a friend prepare for technical networks and systems exams.

Team Submissions: *I built and designed this project entirely as a solo developer.

Thanks for participating! @thepracticaldev

Falcon AI – Intelligent Human Interaction & Agentic Automation System

krithikgokuls — Sat, 15 Aug 2026 15:40:24 +0000

My 10-Day Journey with Murf AI | VoiceForBharat Edition

I’m excited to share my experience completing the 10 Days of Voice Agents — VoiceForBharat Edition challenge by Murf AI.

Over these 10 days, I built and progressively enhanced an AI-powered interaction system using Murf Falcon, the fastest TTS API, along with LiveKit and Python.

The project started with basic real-time interaction and gradually evolved into a system capable of handling memory, tools, outbound communication, human escalation, call analytics, and specialist-agent handoffs.

1. The Problem and the Users

For this project, I selected a Financial Services use case.

Financial services often involve users asking questions, checking eligibility, requesting information, and needing assistance with different processes.

The goal was to explore how an AI-powered interaction system could make these experiences more natural and accessible.

Instead of requiring users to navigate complex interfaces or type every question, a voice-based interaction allows them to communicate naturally.

The AI system is designed to assist users while also understanding its limitations and knowing when to involve a human or another specialised agent.

2. Introducing Falcon AI
Falcon AI – Intelligent Human Interaction & Agentic Automation System

Falcon AI is an AI-powered interaction system designed to demonstrate how intelligent agents can:

Communicate naturally with humans
Remember relevant information
Use tools to perform actions
Initiate outbound interactions
Escalate situations to humans
Analyse conversations
Hand off conversations to specialised agents

The project uses Murf Falcon for voice generation and LiveKit for real-time communication.

3. My 10-Day Journey
Day 1 – Build the Foundation

The journey started by setting up the core real-time AI interaction system.

The goal was to establish the foundation required for users to communicate with the AI.

🎥 Day 1 Demo:
https://youtu.be/09HhFW0SQXI

Day 2 – Personality, Role & Limits

The AI agent was given a defined personality, role, objectives, responsibilities, and safety boundaries.

The Financial Services scenario helped establish clear behaviour and limitations for the agent.

🎥 Day 2 Demo:
https://youtu.be/My_Wc8r5AdQ

Day 3 – Personalised Frontend

The next stage focused on the frontend experience.

I created a personalised interface that allows users to interact with the AI system and observe the current state of the interaction.

🎥 Day 3 Demo:
https://youtu.be/RmWX0Qc_YYo

Day 4 – Persistent Memory

The agent was given persistent memory so that relevant user information could be retained.

This allows returning users to receive more personalised and contextual interactions.

🎥 Day 4 Demo:
https://youtu.be/P-_4OXs09CU

Day 5 – AI Tools & Actions

The agent was connected to tools that allow it to perform useful actions rather than simply generating responses.

For example, the Financial Services use case included functionality related to checking scheme eligibility.

🎥 Day 5 Demo:
https://youtu.be/hbNWKFElsI0

Day 6 – Outbound Calls

The system was extended to support outbound interactions.

Instead of only waiting for users to initiate a conversation, the AI could proactively communicate with users through an outbound calling workflow.

🎥 Day 6 Demo:
https://youtu.be/tVoqUXaY5f0

Day 7 – Human-in-the-Loop

One of the most important capabilities was human escalation.

The AI was designed to identify situations where it should involve a human instead of trying to handle everything independently.

This introduced a practical Human-in-the-Loop workflow.

🎥 Day 7 Demo:
https://youtu.be/cOmmg8e3gic

Day 8 – Call Analytics Dashboard

Once the system could handle interactions, the next step was understanding those interactions.

I built a call analytics dashboard to monitor activity and gain insights into conversations and outcomes.

🎥 Day 8 Demo:
https://youtu.be/Hbqq_rakyHw

Day 9 – Specialist Agent Handoff

The final major capability was specialist-agent handoff.

Instead of expecting one general-purpose agent to handle every situation, the system can transfer a conversation to a specialised agent when another agent is better suited to handle the request.

This introduced the concept of multi-agent collaboration.

🎥 Day 9 Demo:
https://youtu.be/5hWJMglt4Mg

4. How the System Works

The basic architecture can be represented as:

User → Speech-to-Text → AI Agent / LLM → Memory & Tools → Murf Falcon TTS → User

Additional workflows extend the system:

Outbound Calls → Human Escalation → Analytics → Specialist Agent Handoff

Main Components
Speech-to-Text

Converts the user's spoken input into text that the AI system can understand.

LLM / AI Agent

Processes the user's request, maintains conversation context, determines what action is required, and generates a response.

Murf Falcon

Murf Falcon provides the text-to-speech layer, converting AI-generated responses into spoken output.

LiveKit

LiveKit provides the real-time communication infrastructure required for the interaction.

Memory

Stores relevant information so returning users can receive more contextual responses.

Tools

Allow the AI agent to perform specific actions and interact with external functionality.

Human Escalation

Allows the AI to involve a human when a request falls outside its capabilities or requires human assistance.

Specialist Agents

Allow specific tasks to be transferred to agents designed for particular responsibilities.

5. The Difficult Parts

One of the most valuable parts of this challenge was discovering that not everything works perfectly on the first attempt.

Frontend Build Error

While running the frontend locally on localhost, I encountered a build error.

This required troubleshooting the development environment, checking configuration and dependencies, and identifying the issue before continuing with the project.

I have included screenshots/proof of the frontend build error as part of my project documentation.

This was important to me because real-world development is not only about showing the final working application. It is also about understanding and solving the problems encountered during development.

Other Challenges

I also faced difficulties with:

LiveKit configuration
Frontend and backend integration
Environment setup
API configuration
Outbound calling
Connecting multiple components
Runtime debugging
Testing the complete workflow

These challenges helped me improve my troubleshooting and development skills.

*6. What I Learned From the Challenges
*
When something failed, I learned to:

Read the complete error message.
Identify which component caused the problem.
Check dependencies and configuration.
Verify environment variables.
Test individual components separately.
Debug integrations step by step.
Test the complete workflow again.

These practical debugging experiences were one of the most valuable parts of the challenge.

7. Security

When publishing an AI project publicly, protecting sensitive information is extremely important.

I made sure not to expose:

API keys
Secret credentials
Phone numbers
Caller information
Private user data
Sensitive environment variables

API keys should always be stored securely using environment variables or an appropriate secret-management solution.

8. Source Code

The complete project source code is available on GitHub:

🔗 GitHub Repository:
https://github.com/krithikgokuls/ai-voice-agents-murf-ai

The repository contains the implementation and supporting code for the AI interaction system built using Murf AI and LiveKit.

9. What I Learned

The biggest lesson from this challenge is that an AI voice system is much more than a system that can talk.

A useful AI agent should be able to:

Listen → Understand → Remember → Act → Analyse → Escalate → Collaborate

Through this challenge, I gained practical experience with:

Voice AI
Conversational AI
Agentic AI
Persistent memory
Tool calling
Real-time communication
API integration
Telephony
Human-in-the-Loop systems
Call analytics
Multi-agent workflows

10. What I Would Improve Next

There are several areas I would like to explore in future versions:

Better multilingual and code-mixed conversations
More advanced memory management
Improved analytics
Additional specialised agents
Better error handling
Production-ready deployment
Improved authentication
Enhanced frontend experience
Better monitoring and logging
More advanced Human-in-the-Loop workflows

2. Final Thoughts

Completing the 10 Days of Voice Agents — VoiceForBharat Edition challenge was a valuable hands-on learning experience.

The difficult parts were just as important as the successful parts.

The frontend build error, LiveKit configuration, integrations, and debugging challenges taught me that building AI applications requires continuous experimentation, troubleshooting, and problem-solving.

This journey helped me understand how AI agents can move beyond simple conversations and become systems that can interact, remember, act, analyse, escalate, and collaborate.

I’m grateful to Murf AI for providing this opportunity to learn and build with Voice AI.

Project

Falcon AI – Intelligent Human Interaction & Agentic Automation System

Technology Stack

SmartCrowd AI: Intelligent Crowd Management System for Sporting Venues

krithikgokuls — Fri, 10 Apr 2026 10:36:20 +0000

This is a submission for the DEV April Fools Challenge

What I Built

I built SmartCrowd AI, a cloud-based system that tracks crowd density, predicts congestion, and suggests better routes in large venues, making events smoother and safer.

Demo

https://smartcrowd-145365073250.asia-south1.run.app/predict?zone=Food Court
https://smartcrowd-145365073250.asia-south1.run.app/crowd
https://smartcrowd-145365073250.asia-south1.run.app/predict?zone=Gate A

Code

https://github.com/krithikgokuls/SmartCrowd-AI-System-using-Cloud-Run

How I Built It

I built SmartCrowd AI using Python with Flask for the backend API, containerized the application using Docker, and deployed it on Google Cloud Run for scalability. The system provides real-time crowd data and predictions through REST APIs, which can be accessed by a web or mobile interface.

Prize Category

Best Google AI Usage

This project uses Google Cloud services like Cloud Run to build a scalable, real-time system and demonstrates how AI-based predictions can improve crowd management and user experience in large venues.

Experience On Hacktoberfest

krithikgokuls — Sun, 27 Oct 2024 14:37:44 +0000

Introduction to Hacktoberfest:
Hactoberfest is one of the most popular, well - attended annual global Contributions of Open Source.Every October , people from a variety of expertise backgrounds , and experience levels from all over the world come together and use their skills to help maintain open source projects.
Open Source Contributor:
Contributing to open source software is not only a way to share your skills in a particular language or tech stack , it can be rewarding practice to share your engineering knowledge and collaborate with the developer community.
Host of Hactoberfest and Years of Experience:
Hactoberfest is Digital Ocean's annual event that encourages people to contribute to open sourcer throughout october.This month marks the 11th year of hactoberfest , a month long virtual event that encourages developers to contribute to the open source projects that so many of us rely on.

My Experience on Hactoberfest 2024:-

I recently made my first open-source contributions through Hacktoberfest, submitting accepted PRs/MRs with the "hacktoberfest" or "hacktoberfest-accepted" label across five repositories. My contributions involved coding in languages like Python and Java. This experience helped me understand the process of contributing to projects, from creating repositories to working with project maintainers and awaiting the seven-day review period on GitHub and GitLab. Overall, it was an engaging introduction to open source, thanks to Hacktoberfest and DigitalOcean’s support.

digitalocean #hactoberfest #dev #opensource #contributor @digitalocean_staff

Post Done By
KRITHIK GOKUL S
Contact via email id:krithikgokuls@gmail.com