PromptPilot AI

We noticed a massive gap in the current AI landscape: The "Blank Page" Problem. While models like Gemini are incredibly powerful, most users (our friends, family, and non-tech peers) get poor results because they treat AI like a keyword search engine. They don't know how to write chain-of-thought prompts or define personas. We asked ourselves: What if the AI could "read between the lines"? We wanted to build an Orchestration Layer that sits between the user and the raw model—a system that translates "casual human intent" into "professional engineering execution." We were also inspired by the Gen Z aesthetic—we wanted to move away from sterile, corporate chat interfaces and build something that feels alive, neon, and responsive.

Built With

css
css3
diffusers
elevenlabs
fastapi
gemini
google
html5
javascript
marked.js
python
pytorch
tailwind
uvicorn

Submitted to

AI Partner Catalyst: Accelerate Innovation

Created by

Backend Engineering (FastAPI): we built the Python microservice that serves as the brain of the app. We wrote the logic that intercepts casual user queries and "rewrites" them into professional prompts using persona injection.
API Integration: we personally integrated Google Gemini 1.5 Pro for intelligence and ElevenLabs Turbo V2 for ultra-low latency multilingual voice.
The "Failsafe" Architecture: we encountered a major issue where Google model names kept changing (causing 404 errors), so we took help of gemini pro for a custom Smart Model Scanner algorithm that dynamically checks API permissions and selects the best available model before the chat starts.
Frontend Development: we built the entire UI from scratch using HTML5, Tailwind CSS, and Vanilla JavaScript. We avoided heavy frameworks to keep the app lightweight and fast.
What we Most Proud Of:
We are most proud of the Multimodal "Magic"—getting text, voice (ElevenLabs), and vision (Local Stable Diffusion) to all work seamlessly in a single interface without crashing the Python thread loop.
What we Learned:
We learned that "Prompt Engineering" isn't just typing words; it's a programmable logic layer. By treating prompts as code that can be optimized programmatically, we was able to massively improve the quality of the AI's output.
We are will to upgrade Promptpilot AI in Future

Ayyappan Ayyanan
Aaliyah Rehana
Magadheesh

Updates

Ayyappan Ayyanan started this project — Dec 31, 2025 03:22 AM EST

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.