Download Resume
Done. Life logo

Founding Engineer & Product Lead

Zero-to-One AI Startup

Done. Life·Sydney, Australia

Dec 2024Present

Founding RoleCurrentAI/MLMobileServerless
☕☕☕☕☕
Coffee Required
Principal
Complexity
⭐⭐⭐⭐
Documentation
👍👍
Would Do Again

Synopsis

Leading the end-to-end product lifecycle for an AI-first productivity platform that automates 'Life Admin' by converting unstructured emails, PDFs, and images into structured calendar events.

The Making Of

🎯 The Challenge

Users drown in 'Life Admin'—bills, appointments, event invites buried in emails. Manually copying dates to calendars is tedious and error-prone. The AI challenge: LLMs frequently 'hallucinated' dates, calculating relative dates from their training data cut-off instead of the current date.

💡 The Solution

Engineered a 'Context Anchor' injection system that dynamically injects the current ISO timestamp and day of week at inference time. Built a 'Split-Path Engine' to route traffic between low-latency text models and high-compute vision models based on payload analysis. Implemented aggressive 'Junk Filters' to discard non-actionable attachments (<5KB signatures/icons).

🚀 The Impact

Completely solved the temporal hallucination issue. The 'Forward-to-Email' feature became the app's most-used ingress method. Reduced API costs by filtering 90% of junk attachments. Successfully launched on App Store within 8-week timeline.

Key Contributions

  • 1Architected a serverless, event-driven backend to automate complex 'Life Admin' workflows for consumers
  • 2Engineered a custom traffic router that dynamically directs payloads between low-latency text models and high-compute vision models, reducing processing overhead by ~40% and response times by ~70%
  • 3Solved temporal hallucination issues in LLMs by injecting dynamic context anchors and Chain-of-Thought protocols
  • 4Led the end-to-end product lifecycle from Zero-to-One, managing full stack & navigating App Store compliance

Key Metrics

~40%
Processing Error Reduction
90%
Non-actionable Attachment Filtering
8 weeks
Time to App Store

Skills & Technologies

  • AWS LambdaServerless Compute
  • OpenAI GPT-4oAI Engine
  • OpenAI VisionImage Processing
  • Make.comWorkflow Orchestration
  • AdaloMobile Frontend
  • DynamoDBDatabase
  • IAPHUBIn-App Purchases

Quick Info

Company
Done. Life
Location
Sydney, Australia
Type
Founding Role
Duration
8 weeks (Zero-to-One)
Complexity
Principal
← Back to All Experience