Embizzolator

Inspiration

The inspiration for Embizzolator came from a universal experience: the corporate meeting. We've all been in conference rooms where buzzwords and jargon fly so thick and fast that the actual meaning is lost. I wanted to create a tool that was both a satirical commentary on this corporate culture and a genuinely powerful creative aid. Embizzolator started as a joke— "what if I could automate synergy?"—and grew into a fully-featured application that explores the intersection of language, AI, and professional personas. My goal was to build something fun, polished, and technically robust that could make anyone sound like a C-suite executive with the tap of a button.

What it does

Embizzolator is a next-generation, AI-powered communication tool built for Android. It translates plain, everyday English into high-impact, jargon-heavy corporate speak. The user can type or speak a phrase, and the app leverages the powerful OpenAI gpt-oss:20b model via a high-speed Groq API endpoint to generate a "translated" version packed with buzzwords and professional idioms.

The app is highly customizable, offering granular control over the AI's output:

• Sliders: Users can adjust the Jargon Density, Urgency Meter, and Verbosity of the response on a five-point scale from "Low" to "High."
• Corporate Personas: Users can instruct the AI to adopt the voice of a Business Executive, Engineering Manager, Agile Product Owner, or Marketing Executive, each with its unique flavor of jargon.
• Dynamic Theming: The entire app's UI can be changed on the fly by selecting "Branding Guidelines" like "Executive Mahogany" or "Cube Farm Chic," demonstrating a deep integration with the Material You theming system.
• Full Functionality: The app includes speech-to-text for input, text-to-speech for playback, and secure, password-protected storage for API credentials.

How we built it

Embizzolator is an Android native application built entirely in Kotlin using the modern Jetpack Compose declarative UI toolkit.

• Frontend & UI: The entire user interface, including navigation between the main screen and the settings dashboard, is built with Jetpack Compose. State management is handled by a Jetpack ViewModel and StateFlow, ensuring a reactive and lifecycle-aware UI.
• AI & Backend: We integrated the OpenAI gpt-oss:20b model via the Groq API. We chose Groq for its incredibly high-speed inference, which makes the app feel responsive and real-time.
• Networking: All API calls are handled using OkHttp, a robust and industry-standard HTTP client. Kotlinx.serialization is used for parsing the JSON request and response payloads.
• Security: To protect user credentials, we implemented the Jetpack Security library, using EncryptedSharedPreferences to store the API URL, key, and an optional user-set password on the device.
• Native Features: We used Android's built-in TextToSpeech engine for audio playback and the RecognizerIntent for speech-to-text input.

Challenges we ran into

Implementing the dynamic theming system was a significant challenge. It required creating multiple ColorScheme objects and ensuring that every component in the app correctly referenced MaterialTheme.colorScheme instead of hardcoded colors. Another challenge was prompt engineering; creating a single, robust prompt that correctly instructed the LLM to respond to five different dynamic parameters (user input, jargon, urgency, verbosity, and persona) took considerable iteration and testing to get right. Finally, managing the lifecycle of the TextToSpeech engine and handling secure storage in a way that was both safe and user-friendly required careful planning.

Accomplishments that we're proud of

I am proud of building a polished, end-to-end Android application that is not only functional but also fun to use. Successfully integrating a high-speed LLM API to perform a creative and complex task was a major achievement. I am particularly proud of the highly customizable user experience, from the dynamic prompt parameters that give users fine-grained control over the AI to the fully dynamic theming system that alters the app's entire appearance. Implementing on-device security for user credentials with a lock/unlock workflow was a great learning experience that adds a professional touch to the app.

What we learned

This project was a deep dive into modern Android development. I learned the best practices for managing complex, screen-level state with a ViewModel and how to create a flexible, dynamic UI with Jetpack Compose. We also gained valuable experience in securely storing sensitive data on-device with EncryptedSharedPreferences and in the art of prompt engineering to achieve precise control over an LLM's output. It was a fantastic exercise in connecting a powerful backend service to a native mobile frontend.

What's next for Embizzolator

The roadmap for Embizzolator is full of potential synergies. Future sprints could include:

• User Presets: Allow users to save their favorite combinations of slider and dropdown settings for quick access.
• Share Functionality: Integrate with Android's share sheet to easily send generated jargon to Slack, email, or social media.
• Expanded Content: Add more corporate personas (e.g., "Venture Capitalist," "HR Business Partner") and more branding themes.
• "De-bizzolator" Mode: A reverse-translation feature to turn dense corporate jargon back into plain English.

NB: The current Appetize.io demo assumes you have credentials for the Groq.com endpoint for accessing this model.

Built With

• kotlin