Prasun
AI Product Specialist. I can build and deploy ANY AI product
New to Contra
Prasun is ready for their next project!
Hello everyone, My name is Prasun and this is AURA.
A BRIEF DESCRIPTION OF THE PROJECT " AURA " WE BUILT-
AURA is an interactive, luxury sensory-cognitive wellness workspace designed to reduce stress and foster deep flow states. Moving away from traditional, static software grids, AURA implements an HTML5 canvas vector flow field running at a locked 60 FPS. This interactive backdrop translates user mouse trajectories dynamically into fluid particle paths while feeding spatial data into a multi-voice synthesizer built directly on the native Web Audio API. The result is a living user interface where touch vectors create real-time, resonance-filtered ambient chords and evolving visual physics.
1. Link to project :- https://38425a77-aura-cognitive-flow-space.netlify.app/
2. Optional backup link to the editable Stitch workspace: https://stitch.withgoogle.com/projects/8647869648563302522
3. Brief explanation of how we used Stitch:
Google Stitch was the main control center for the whole project. We started by writing down all our styling rules—like our specific dark blue colors, see-through panels, and smooth 24-pixel corners—in a file named DESIGN.md (http://DESIGN.md). Because Stitch reads this file as a strict visual guide, its AI engine automatically placed all our volume sliders and screen controls perfectly without breaking our theme. For the advanced setup, we used a terminal tool called an MCP server to link our online design directly to our computer. This let our offline coding tools inside Google Antigravity 2.0 scan our project files, check our work, and make sure our music audio settings were perfectly synced.
4. Feedback -
The Good Parts: The DESIGN.md (http://DESIGN.md) file system is amazing. It completely stops the AI from changing your colors or fonts randomly on later edits, keeping the design looking clean and consistent. Also, the one-click Netlify export button makes publishing incredibly fast, letting you launch a live website in seconds without setting up complicated servers.
The Bottlenecks: Right now, if you manually tweak a design element directly on the canvas, those changes do not automatically update inside the text of your DESIGN.md (http://DESIGN.md) file.
My Suggestions:
Make it a two-way street: if we change a color visually on the screen, have Stitch rewrite that color token in the DESIGN.md (http://DESIGN.md) file automatically.
Add a "lock" button to elements on the canvas so that when the AI edits a specific box, it doesn't accidentally shift the positions of neighboring elements.
Build a simple visual menu for the local MCP server link so designers can connect their computer code without needing to type lines in a terminal.
5. A FULL WALKTHROUGH OF THE PROCESS : -
2
1
58
Context is King. 👑 I just stress-tested Gemini 3.1 Pro’s 2-million-token window against the EU AI Act.
Most RAG systems suffer from "Long-Range Dependency"—they forget what happened on page 5 by the time they hit page 200. I built a Technical Synthesis Engine that eliminates this blind spot.
What I did:
Engine: Google AI Studio (Gemini 3.1 Pro)
Logic: Temperature 0.0 for zero-hallucination accuracy.
Input: An almost 200 page technical manual.
The Result: A searchable policy assistant that cross-references sections with 100% citation accuracy. It found a "needle-in-a-haystack" regulatory penalty on page 142 that basic summaries missed.
0
7
How I compressed a 40-hour Real Estate Research cycle into a 12-minute AI Infrastructure.
Manual market research and asset creation are the biggest silent killers of brokerage margins. I’ve engineered a localized AI pipeline specifically for the Dubai/Abu Dhabi real estate markets that automates the "Research-to-Asset" lifecycle.
1) Live Data Ingestion: Utilizing Perplexity AI Pro to scrape real-time 2026 transactional data, rental yields (ROI), and infrastructure developments (e.g., Palm Jebel Ali Q2 absorption rates).
2) Contextual Synthesis: Deploying Gemini 3.1 Pro to process raw data into high-converting investor narratives and LinkedIn hooks.
3) Automated Creative Output: Direct mapping of synthesized data into Adobe Express Pro for branded, high-fidelity carousels and investor dossiers.
The Result: High-authority digital assets for top-producing agents delivered in minutes, not days. This system removes the "admin-bottleneck," allowing brokers to focus strictly on viewings and closing deals.
0
14
Can AI replace a $200k/year Investment Analyst? I built an Executive Intelligence Dashboard to find out.
As a part of my first solo project as an AI Productivity Specialist, I just open-sourced my AI Investment Research Framework on GitHub. It shows the exact Chain-of-Thought prompts I used with Gemini 3.1 Pro and Perplexity AI Pro to stress-test market data for a $15M VC scenario.
The Client requested to know specifically, if startups focusing on 'Autonomous Freight Auditing' have a real 'moat' or if they will be crushed by giants like SAP and Oracle in the next 18 months.
The Goal of my Project: Providing a data-backed "Investment Memo" that proves one can use AI to synthesize massive amounts of contradictory market data into a clear "Invest" or "Pass" recommendation.
Here is the finalised content done with Adobe Express.
0
18
Made and used a customised version of chatterbox multilingual with unlimited words to use the TTS model for AI based voice cloning. In this video recording I show the exact set of steps laying out how it is to be pulled off along with a sample at the end of the clip showing the output .
0
29