Home/ Team/ Son Le

Platform Engineer · EMintel

Son Le

Platform engineer behind EMintel, Earned Media's proprietary intelligence platform that synthesises live LLM probe data, traditional search signals, and internal comms into a single decision layer. Computer Science graduate from the University of Wollongong (Artificial Intelligence & Big Data, WAM 84.5). Founder of CobbyIQ — an AI-driven SaaS built end-to-end on Claude API, Voyage AI, and Pinecone.

About Son

Son leads engineering on EMintel, the proprietary intelligence layer that powers every Earned Media client engagement. He architects the data pipelines, AI integrations, and execution surfaces that turn fragmented inputs — Ahrefs, Semrush, GSC, GA4, Fireflies, Slack, and live LLM probes across ChatGPT, Gemini, Perplexity, Claude and Google AI — into a single, queryable intelligence system.

Before joining Earned Media, Son founded CobbyIQ, an AI-driven SaaS where he architected a multi-service integration layer orchestrating Claude API, Voyage AI, and Pinecone — handling authentication, error recovery, retrieval-augmented generation, and document ingestion across a multi-tenant FastAPI backend. He also founded VaniaFlorist, a production e-commerce platform on React, TypeScript, Supabase and Stripe with webhook-driven order fulfilment. Earlier, as an intern at DigiWize / SampleAssist, he built an automated pipeline that pulled GitHub events into a PostgreSQL store and surfaced DORA performance metrics across teams — cutting query latency by 35%.

Son holds a Bachelor of Computer Science from the University of Wollongong, specialising in Artificial Intelligence and Big Data, graduating with a WAM of 84.5.

What Son builds at Earned Media

Every pillar of Earned Media's service delivery sits on top of EMintel, and Son is the engineer turning that architecture into shipped infrastructure:

  • Live LLM probe pipelines. Continuous crawls of ChatGPT, Gemini, Perplexity, Claude, and Google AI/AIO — capturing mention and citation data at the query level.
  • Client intelligence profiles. A 10-section "Ground Truth" embedding store anchoring every evaluation in client-specific context.
  • The 4-rubric evaluator. Powered by Claude Sonnet, scoring every AI response on confidence, completeness, visibility, and sentiment.
  • Agent skills dispatcher. Autonomous routing with progressive disclosure — the right tool, the right context, the right time.
  • Bidirectional task sync. Approved strategies pushed straight to the execution team's task boards. Friday team scores, Monday performance briefs, zero busywork.

Selected projects

CobbyIQ — Founder & Full-Stack AI Developer (2025 – present)
AI-driven SaaS built end-to-end on Claude API, Voyage AI, and Pinecone. FastAPI backend with multi-tenant session management, PostgreSQL persistence, and structured JSON exchange. RAG pipeline normalising unstructured documents into vector embeddings for context-aware retrieval. Deployed on Vercel with full CI/CD.
VaniaFlorist — Founder & Full-Stack Developer (2024 – present)
Production e-commerce platform on React, TypeScript, and Supabase (PostgreSQL). Stripe payment processing with webhook-driven order updates, Resend API for transactional email, and on-page SEO baked in.
DigiWize / SampleAssist — Software Engineering Intern (2024 – 2025)
Automated data pipeline integrating GitHub REST API with PostgreSQL — transforming raw event data into DORA performance metrics (deployment frequency, lead time, change failure rate). Optimised SQL queries for a 35% latency reduction. Built OAuth flows for secure third-party API access.
One-shot Facial Recognition — University Project (2024)
Siamese CNN trained with contrastive loss for one-shot identification, integrated into a broader system with LSTM-based wake-word detection. Built in Python and TensorFlow.

Get in touch

If you've got a question about how EMintel works, the architecture choices behind the dual-engine platform, or want to talk through how AI tooling can sit underneath an SEO/GEO agency — drop Son a line at son@earnedmedia.com.au, or book a call here.

Want to see EMintel in action?

30 minutes. We'll walk you through how the platform actually works, the data sources behind it, and what it surfaces for clients each week.

Book a call