Sebastian Tirelli

I work on LLM systems in production — observability, cost efficiency, evaluation. Previously led engineering teams, built products end-to-end, and architected systems at scale. 12 years in the craft.

Writing

Longer posts on LLM systems infrastructure, production cost optimization, and RAG engineering. Publication schedule: irregular, when there's something worth reading.

First post landing soon.

Selected projects

• llamaindex-embedding-churn
  Reproducer for a two-bug interaction in LlamaIndex that silently wastes embedding spend under specific ingestion patterns. Five progressive experiments — hash comparison, counting embedder, real OpenAI API, reader-format families, S3 end-to-end.
• ChatGPT LightSession
  Chrome extension with 60,000+ users (4.8★) that trims long ChatGPT threads client-side, keeping only the last N messages along the active path. Fixes slow/laggy sessions on chatgpt.com.

Background

12 years in software engineering. 4 years leading engineering teams of 35+. Built an AI SDR platform end-to-end as a solo founder (LangGraph, async workers, human-in-the-loop). Previously led architecture and evaluation / benchmarking pipelines for an AI memory and retrieval platform. Two postgraduate certificates from UC Berkeley — Machine Learning & AI, and Technology Leadership.

Based in Argentina. Italian + Argentine citizenship — available in LATAM, EU, and remote worldwide.

Elsewhere

• GitHub
• LinkedIn
• X / Twitter
• Email