PodMorph

Overview

I led the development of a cross-platform application that uses AI to transform podcasts and radio programmes into video content for social media. I built it from scratch across 7 platforms. The most important part wasn't the AI - it was making sure the AI didn't make things up.

A collaboration between Captain McFunshine's Software Engineering, Overcoat Media, an independent project leader and an independent designer, 2024.

Technical Achievements

Software Architecture & Cross-Platform Development

As Lead Architect, I designed and implemented a comprehensive cross-platform solution:

• Architected and implemented a comprehensive Kotlin Multiplatform (KMP) solution delivering native applications for 7 platforms (iOS, Android, Windows, macOS on both x64 and ARM, Linux on both x64 and ARM)
• We collaboratively designed and implemented the UI using Compose Multiplatform, enabling consistent interfaces across all supported platforms
• Established a robust Model-View-Intent (MVI) architecture for predictable state management and UI interactions
• Implemented client-server architecture to control AI interaction and balance processing requirements between local devices and cloud resources
• Integrated multiple AI services with fallback mechanisms to ensure service continuity

Cloud Infrastructure & API Integration

I developed a robust cloud infrastructure to handle complex media processing:

• Designed cloud-based processing workflow using Azure Container Instances and AWS S3 for data storage
• Implemented PostgreSQL database management with data models for user accounts, projects, payments, and content artifacts
• Orchestrated multiple AI services including Groq, Anthropic Claude, OpenAI, Fal.ai, Whisper, and Google Vertex AI with fallback mechanisms
• Built optimized media processing pipelines for transcription, summarization, and text, image and video content generation
• Created secure API communication using Ktor with proper authentication and error handling

Media Processing & Transformation

The heart of PodMorph lies in its sophisticated media processing capabilities:

• Developed advanced text-to-image generation workflows using multiple AI models (DALL-E, Flux, Idiogram) for diverse creative outputs
• Created sophisticated audio/video processing capabilities using FFMPEG for precise clip extraction and multi-layer composition
• Implemented intelligent text overlay systems with dynamic sizing and multi-layer outlining for readability
• Built systems to automatically synchronize text animations with audio using segment timings
• Integrated natural language processing for entity recognition, sentiment analysis, and quote extraction

Key Contributions

• Responsible for architecture and development of the product
• Created a complete AI processing pipeline from audio transcription to fully-rendered social media content
• Implemented intelligent content transformation with multiple specialized outputs (Twitter threads, Instagram Reels, LinkedIn posts, etc.)
• Developed sophisticated content verification algorithms to ensure AI-generated content matches source material
• Built a consistent cross-platform user experience that feels native on each supported operating system
• Designed intelligent fallback mechanisms throughout the system to ensure reliability during API outages

Technologies & Skills

• Kotlin Multiplatform (KMP) for cross-platform development
• Compose Multiplatform for UI framework
• Model-View-Intent (MVI) architecture pattern
• Azure Container Instances and AWS S3 for cloud infrastructure
• PostgreSQL database and Exposed ORM
• AI service integration (Groq, Anthropic Claude, OpenAI, Google Vertex)
• FFMPEG for media manipulation
• Ktor for networking and API communication
• HikariCP for database connection pooling
• CI/CD with GitHub Actions
• Docker containerization