MeDraft - Self-Discovery through social platforms

Most career tools ask who you think you are; social platforms reveal who you really are. MeDraft bridges that gap by using RIASEC personality quiz and your Youtube Takeout

Comment

MeDraft – Self‑Discovery through Social Platforms

A one‑month hackathon project that turns your digital footprint into a mirror for personal growth.

Time to Rediscover Ourselves

When my wife decided to leave her hospital job after years of exhausting night shifts, she realised she no longer knew what truly motivated her.

The feeling echoed my own post‑university uncertainty: we go from classroom to workplace on autopilot, rarely pausing to ask where we really want to go.

We envisioned a short “stop‑over” that blended honest self‑reflection with the digital breadcrumbs we leave every day. That spark became MeDraft.

What it does

MeDraft is a self‑assessment SaaS that fuses two complementary profiles:

Profile Data Source Insight Gained
Conscious Validated psychological surveys such as the RIASEC Holland Codes How you describe yourself
Unconscious Behaviour captured in social‑media take‑outs (YouTube, X/Twitter …) How you actually behave

By weaving these lenses together we deliver an actionable report that reveals blind spots between intention and action – plus concrete next steps.

Next milestone: Build a Retrieval‑Augmented‑Generation (RAG) chat agent so users can talk to their unconscious profile in natural language.

Current State of Development

1. Market Gap Analysis

  • Pain #1 – Surveys are redundant & overlong.
  • Pain #2 – Traditional tools only tell you what you already know; there’s little genuine self‑discovery.

Our MVP attacks both problems head‑on.

2. Tech Stack at a Glance

Layer Tooling Why This Choice?
Frontend React + modular CSS Rapid, component‑driven UI; full styling control
Auth & Data Supabase (auth, DB, storage, edge functions) One‑stop scalable backend
Jobs Supabase cron + external Netlify Function Off‑load heavy queues & contact‑form webhooks

3. Goodbye Boring Surveys

  • Custom survey engine stored as: SurveyTemplate → SurveyQuestion → SurveyAnswer → SurveyResponse
  • Admin panel lets us upload/edit JSON definitions on the fly.
  • Users can retake surveys; the latest completion overwrites previous answers.

4. Privacy‑First Processing

  1. Upload ZIP ➜ temp bucket ➜ edge function ➜ delete raw ZIP.
  2. We currently ingest the last 1 000 videos & 250 subscriptions per YouTube take‑out – enough to sketch tastes without over‑processing.
  3. A cron worker fetches missing metadata (title, tags, etc.) in batches of 50 to respect YouTube API limits & Supabase 10‑second edge runtime.
  4. Users can delete their take‑out any time, which cascades all derived data.

5. First‑Pass AI Analysis

Time was tight, so instead of a local model we call the OpenAI API from a Supabase edge function. Even this lightweight approach surfaces fresh, actionable insights – e.g., it nudged me toward artistic hobbies I’d only recently started to explore.

⚔️ Challenges Faced

Social‑Platform APIs Are… Tough

Public endpoints are incomplete, pricey, or both. Requiring user take‑outs is a pragmatic but clunky workaround.

Take‑outs Are Messy

Every platform exports data differently; even YouTube’s ZIP is huge. Normalising formats is an ongoing battle.

Getting the Word Out

Great feedback demands users, but user acquisition steals time from building. Friends & early adopters have been kind, yet we know it’s still an MVP.

🏆 Accomplishments

  • Shipped a functional MVP in one month – a personal focus record!
  • Learnt Supabase inside‑out (auth, cron, edge functions).
  • Proved that small datasets (<1 000 items) already yield meaningful self‑insights.

📚 What I Learned

  • "Take‑out" was a new concept; now I grasp how companies really store our data.
  • Backend‑first development pairs beautifully with AI‑generated front‑ends (thanks, Bolt!).
  • Supabase’s developer experience is stellar.
  • Most importantly: building tools that help people is what keeps me motivated.

📌 Next Steps

  1. Ship the RAG chat mentor on a self‑hosted LLM.
  2. Expand beyond YouTube – each new platform will get its own parser.
  3. Layer on habit‑tracking to turn insights into action.

Feedback welcome!
Ping me on X (@mintoolkit00)

Built With

Share this project:

Updates