Overview
Markd is a social marketing automation tool that uses AI to generate captions, hashtags, and content plans. The platform helps creators and small businesses maintain consistent social presence without spending hours on content creation.
Built for people who know they need to post but hate staring at a blank screen.
My Role
As lead software engineer and project manager, I built Markd solo:
- Product Vision — Designed AI-powered content workflow
- Full-Stack Development — Built the complete Next.js platform
- AI Integration — Implemented OpenAI API with custom prompts
- Content Calendar — Created scheduling and planning features
- UX Design — Designed intuitive interface for non-technical users
Built entirely by me.
Architecture
AI-native architecture designed for content creation:
Backend
Next.js API routes with TypeScript. Prisma for PostgreSQL. OpenAI API for content generation.
AI Layer
Custom prompt engineering for different platforms. Context-aware generation using brand voice. Rate limiting and cost management.
Content Management
Calendar view with drag-and-drop. Draft system with version history. Approval workflow for teams.
Analytics
Post performance tracking. Content type analysis. Optimal timing recommendations.
Hard Problems
1. Brand Voice Consistency
AI tends toward generic output. Built brand profile system that fine-tunes prompts for consistent voice.
2. Platform-Specific Optimization
Twitter differs from Instagram differs from LinkedIn. Created platform-aware generation with appropriate formatting.
3. Hashtag Relevance
Trending hashtags change constantly. Implemented dynamic hashtag research with relevance scoring.
4. Generation Quality
Sometimes AI output misses the mark. Added iteration system where users can guide regeneration with feedback.
What I'd Improve Today
- Add image generation with DALL-E or Stable Diffusion
- Implement direct posting to social platforms
- Create AI-powered engagement responses
- Add competitor analysis for content inspiration
- Build team collaboration features
- Implement A/B testing for caption variations