Skip to content
View stirelli's full-sized avatar
10 followers · 12 following
  • Argentina

Achievements

Achievement: Pair Extraordinairex3Achievement: YOLOAchievement: QuickdrawAchievement: Arctic Code Vault ContributorAchievement: Pull Sharkx3

Achievements

Achievement: Pair Extraordinairex3Achievement: YOLOAchievement: QuickdrawAchievement: Arctic Code Vault ContributorAchievement: Pull Sharkx3

Highlights

  • Pro

Block or report stirelli

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
stirelli/README.md

Sebastian Tirelli

Twelve years building software. Four of those leading engineering teams of 35+. Lately I write up what I find interesting. Rhythm is irregular, only when something is worth reading.

Selected projects

  • llamaindex-embedding-churn. Reproducer for a two-bug interaction in llama-index-core that silently re-embeds unchanged content across half the fsspec backend ecosystem. Three-line fix filed upstream as PR #21462. Writeup.
  • ChatGPT LightSession. Chrome extension that trims long ChatGPT threads client-side. Currently at 60,000+ users and 4.8 stars, shipped without paid distribution. Writeup.
  • CodeGraph. Local MCP server I built for Claude Code. Pre-computes a tree-sitter call graph and exposes it through six tools. On a 484-file FastAPI benchmark: 59% fewer investigation tokens, 60% fewer turns, and 82 seconds less wall time, with file recall preserved at 100%. Writeup.

Background

  • 12 years software engineering. 4 of those leading teams of 35+.
  • Built an AI SDR platform end-to-end as a solo founder (LangGraph, async workers, human-in-the-loop).
  • Worked on a hybrid retrieval stack (vector + graph + keyword) and built its benchmarking pipeline, scored against LongMemEval and LoCoMo.
  • UC Berkeley postgrad certificates in Machine Learning and AI, and Technology Leadership.

Elsewhere

sebastiantirelli.com · LinkedIn · X / Twitter · tirelli@gmail.com

Pinned Loading

  1. agentheon-overview agentheon-overview Public

    AI SDR platform using LLM agents to automate outbound workflows, with focus on personalization, deliverability, and system design.

  2. federated-churn-ai federated-churn-ai Public

    Jupyter Notebook 1

  3. bank-marketing-model-comparison bank-marketing-model-comparison Public

    Jupyter Notebook

  4. used-car-price-prediction-crispdm used-car-price-prediction-crispdm Public

    Analysis and prediction of used car prices using machine learning models following the CRISP-DM framework.

    Jupyter Notebook

  5. amazon-coupon-analysis amazon-coupon-analysis Public

    Jupyter Notebook 1