Work
AI & Automation·2025
Document-Flow Automator
AI onboarding platform for financial ops - document validation and summary generation.
Role
Builder (internal)
Stack
FlaskPythonn8nGPT-4GeminiPostgreSQLTwilioZoho
Impact
40% faster reviews · 35% less manual reporting
Problem
Multi-team document onboarding was a manual bottleneck - slow reviews, no QA backlog visibility, and inconsistent handoffs.
Architecture
Flask backend · n8n orchestration · GPT-4 + Gemini for validation and summary · Zoho + Twilio integration · PostgreSQL audit log.
Key decisions
- Pair LLM summarization with rule-based validation so the system never blindly trusts an AI output.
- Build the audit log first - compliance was a hard requirement, not an afterthought.
- Make the human handoff explicit instead of hiding it behind the AI.
Outcome
40% faster reviews · zero QA backlog · 35% less manual reporting time.
What I learned
AI workflows land or die based on the seams between model and human - get the seams right first.
What's next
Template the system for adjacent onboarding flows.
Next
Browse more selected work