Experience

On the ground,
in production.

Not a sandbox project — live systems, real users, real billing pipelines on the line.

Iris
AI/ML Engineering Intern
MAR 2026 — PRESENT
  • Designed and deployed an LLM-powered agentic template generation pipeline with automated reasoning loops and memoised workflow decomposition, cutting project setup latency by 70% and eliminating manual configuration for 100% of new user onboarding flows.
  • Built a multi-LLM orchestration layer routing across Claude, GPT-4, and Gemini with capability-based dispatch, dynamic context management, and intelligent load balancing — serving real production traffic.
  • Traced a critical billing infrastructure bug through payment-flow logs back to an off-by-one error in the credits ledger; patched and regression-tested, restoring the full monetisation pipeline within 24 hours.
Context

What Iris is.

About
EARLY-STAGE AI STARTUP
  • Iris is a WhatsApp-native AI product — orchestration, onboarding, and billing all run in production with real users on the other end.
  • Working at this stage means owning a system end-to-end: design the architecture, ship it, then debug it when something breaks in prod at 2am.
  • That ownership is the throughline across the LLM router, the onboarding pipeline, and the ledger fix below.