Home

About Me

View My CV

Home

About Me

View My CV

Home

About Me

View My CV

UP There

"Up There" began as a high-stakes challenge during SCAD’s Fluxathon: build a unique weather application in 24 hours. When traditional design-to-code methods failed critically just hours before the deadline, I pivoted the entire strategy to Vibe Coding. By orchestrating a team of specialized AI agents within Google’s AntiGravity ecosystem, I transformed a stalled prototype into a fully functional, live-deployed space weather platform. This project serves as a proof-of-concept for AI-Augmented Engineering, demonstrating how a single Design Manager can orchestrate a synthetic workforce to deliver full-stack application at speed.

View Case Study

Project Type

Designathon Project

Timeline

72 Hrs

Skills & Methods

Vibe-Coding

Agentic AI Workflow

Figma UI Design

Google- Anti Gravity

API Integration

Multi-Agent Systems

Prompt Engineering

Rapid Prototyping

Git Version Control

Change Management

AI Orchestration

Team Leadership

Adaptive Thinking

Clear Articulation

Claude Code

Team

Prathamesh P | Krishnika K | Pranay P

PROBLEM

During "Fluxathon," my team aimed to build "Up There," a comprehensive weather app combining Earth data with Space weather (auroras, meteor showers, eclipses) using NASA APIs.

The Block: Our initial "Design-First" approach failed catastrophically at 4:00 AM on submission day. We had high-fidelity screens, but Claude Code could not stitch the complex UI with the functional logic via command line, leaving us with a broken product and almost zero time.

During "Fluxathon," my team aimed to build "Up There," a comprehensive weather app combining Earth data with Space weather (auroras, meteor showers, eclipses) using NASA APIs.

The Block: Our initial "Design-First" approach failed catastrophically at 4:00 AM on submission day. We had high-fidelity screens, but Claude Code could not stitch the complex UI with the functional logic via command line, leaving us with a broken product and almost zero time.

During "Fluxathon," my team aimed to build "Up There," a comprehensive weather app combining Earth data with Space weather (auroras, meteor showers, eclipses) using NASA APIs.

The Block: Our initial "Design-First" approach failed catastrophically at 4:00 AM on submission day. We had high-fidelity screens, but Claude Code could not stitch the complex UI with the functional logic via command line, leaving us with a broken product and almost zero time.

SOLUTION

I adopted the "Vibe Coding" methodology, leveraging Google AntiGravity to orchestrate a synthetic team of specialized AI Agents. Instead of giving up on this project, I acted as the "Chief of Staff," defining the System Requirement Document (SRD) and managing four distinct agents:

Project Manager Agent: Maintained the "Source of Truth," managed change logs, and ensured alignment with the SRD.
Solution Architect Agent: Planned the tech stack, handled debugging strategies, and ensured scalable architecture.
Front-End Artisan Agent: Equipped with a custom skills.md file, this agent translated wireframes into responsive UI/UX which matched our initial designs from Figma.
Back-End Engineer Agent: Managed complex API handshakes with NASA and Earth weather API services.

Project Manager Agent: Maintained the "Source of Truth," managed change logs, and ensured alignment with the SRD.
Solution Architect Agent: Planned the tech stack, handled debugging strategies, and ensured scalable architecture.
Front-End Artisan Agent: Equipped with a custom skills.md file, this agent translated wireframes into responsive UI/UX which matched our initial designs from Figma.
Back-End Engineer Agent: Managed complex API handshakes with NASA and Earth weather API services.

Project Manager Agent: Maintained the "Source of Truth," managed change logs, and ensured alignment with the SRD.
Solution Architect Agent: Planned the tech stack, handled debugging strategies, and ensured scalable architecture.
Front-End Artisan Agent: Equipped with a custom skills.md file, this agent translated wireframes into responsive UI/UX which matched our initial designs from Figma.
Back-End Engineer Agent: Managed complex API handshakes with NASA and Earth weather API services.

VALUE DELIVERED

0-to-1 Deployment in <12 Hours: Successfully built, debugged, and deployed the fully functional app to Vercel before the deadline.

Proof of Agentic Workflow: Demonstrated that a single "AI Orchestrator" can manage a full-stack software lifecycle by defining clear roles and context (SRD) rather than writing syntax.

Scalable Architecture: The modular agent structure allowed for rapid iteration, proving the viability of "Vibe Coding" for complex, API-heavy applications.

0-to-1 Deployment in <12 Hours: Successfully built, debugged, and deployed the fully functional app to Vercel before the deadline.

Proof of Agentic Workflow: Demonstrated that a single "AI Orchestrator" can manage a full-stack software lifecycle by defining clear roles and context (SRD) rather than writing syntax.

Scalable Architecture: The modular agent structure allowed for rapid iteration, proving the viability of "Vibe Coding" for complex, API-heavy applications.

0-to-1 Deployment in <12 Hours: Successfully built, debugged, and deployed the fully functional app to Vercel before the deadline.

Proof of Agentic Workflow: Demonstrated that a single "AI Orchestrator" can manage a full-stack software lifecycle by defining clear roles and context (SRD) rather than writing syntax.

Scalable Architecture: The modular agent structure allowed for rapid iteration, proving the viability of "Vibe Coding" for complex, API-heavy applications.

APPROACH

Define (The SRD): Wrote a comprehensive System Requirement Document outlining every feature, asset, and data flow to ground the AI models.
Assign (The Roles): Instantiated specific agents in AntiGravity, assigning them distinct "Personas" and responsibilities to prevent context drift.
Equip (The Skills): Created and injected skills.md files to give the Front-End agent specific design constraints and best practices.
Execute (The Build): Iteratively prompted the agents, having the "Architect" review the "Engineer's" code before committing to GitHub.
Deploy: Automated the hosting pipeline via Vercel for immediate live access.

Define (The SRD): Wrote a comprehensive System Requirement Document outlining every feature, asset, and data flow to ground the AI models.
Assign (The Roles): Instantiated specific agents in AntiGravity, assigning them distinct "Personas" and responsibilities to prevent context drift.
Equip (The Skills): Created and injected skills.md files to give the Front-End agent specific design constraints and best practices.
Execute (The Build): Iteratively prompted the agents, having the "Architect" review the "Engineer's" code before committing to GitHub.
Deploy: Automated the hosting pipeline via Vercel for immediate live access.

Define (The SRD): Wrote a comprehensive System Requirement Document outlining every feature, asset, and data flow to ground the AI models.
Assign (The Roles): Instantiated specific agents in AntiGravity, assigning them distinct "Personas" and responsibilities to prevent context drift.
Equip (The Skills): Created and injected skills.md files to give the Front-End agent specific design constraints and best practices.
Execute (The Build): Iteratively prompted the agents, having the "Architect" review the "Engineer's" code before committing to GitHub.
Deploy: Automated the hosting pipeline via Vercel for immediate live access.

SUMMARY

"Up There" was a masterclass in Vibe Coding. It shifted my role from a designer/developer to an Orchestrator of Intelligence. By treating AI not as a tool but as a staff, I turned a failed project into a deployed product in hours, showcasing the future of rapid prototyping and application development.