Hi, this is

Tianshu Fu

AI-driven Engineer

Tianshu Fu at Scale Conference

Hi — I’m Tianshu (Tony) Fu, an AI-driven Data Engineer at Meta.
I design and deploy production-scale LLM tools, including:

MyRA – Agentic Reporting Platform – built on Claude Code (Claude Agent SDK), it automates recurring analyst reporting end to end, cutting reporting time by ~60% for 40+ analysts and executives across Reality Labs and powering 20+ automated reports for Ray-Ban Meta, Oakley, Quest, and Horizon.
Product Social Listening – a ChatGPT-style system that transforms millions of user comments into real-time product insights for Meta Quest and Ray-Ban. It powered real-time sentiment awareness during Meta Connect 2024. [Blog]
LLM Taxonomy Assistant – automatically labels datasets and identifies coverage gaps, saving 2–3 weeks of ML experimentation time per half.

Previously, I was the founding engineer at Knowtex, where I built the first version of a RAG-powered voice AI platform that converts doctor-patient conversations into structured SOAP notes, helping reduce clinician burnout. [App]

  • Name Tianshu Fu
  • Location Los Angeles
  • E-mail tf2502@columbia.edu
  • Phone (646) 898 8948
Download Resume

  • Education

  • Master of Statistics - Columbia University

    Sep 2021 - Dec 2022

    Graduate study in statistics and quantitative analysis. GPA: 3.76.

  • Exchange Student in Data Science - UC Berkeley

    Aug 2018 - Dec 2018

    Coursework: CS100 and Stat 140. GPA: 3.5.

  • Bachelor of Computer Science - Central China Normal University

    Sep 2016 - Jun 2020

    Coursework: Data Structures, Algorithms, and Databases. GPA: 86.7/100.

  • Experience

  • Data Engineer – Meta

    Nov 2023 – Present · Los Angeles, CA

    Lead engineer for LLM-powered products at Reality Labs: MyRA, an agentic reporting platform built on Claude Code (Claude Agent SDK) that cut recurring analyst reporting time by ~60% for 40+ analysts and executives, powering 20+ automated reports across Ray-Ban Meta, Oakley, Quest, and Horizon; a real-time Product Social Listening platform (3K MAU, 10+ orgs) that powered sentiment awareness during Meta Connect 2024; and an LLM Taxonomy Assistant that saves 2–3 weeks of ML experimentation time per half.

    Also build sales data pipelines and dashboards for Ray-Ban Meta smart glasses, plus customer-feedback infrastructure spanning multiple third-party sources. Rebuilt the infrastructure after a provider discontinued service, restoring full coverage while reducing annual costs from approximately $200,000 to under $4,000.

  • Founding Software Engineer – Knowtex (YC S22)

    Jan 2023 – Nov 2023 · San Francisco, CA

    Built a RAG-powered voice AI platform that translates doctor–patient conversations into personalized SOAP notes, cutting documentation time by 70% and boosting note quality by 20%. Fine-tuned an open-source LLM on production medical transcription data, and developed backend services plus a React Native iOS/Android app with secure transcript storage.

  • Data Engineer Intern – Meta

    May 2022 – Aug 2022 · New York, NY

    Built data pipelines and a dashboard that tracked seller retention for Facebook Marketplace, helping product managers plan strategies to grow the seller community.

Skills

Languages

Python, Hack, SQL, Go, Java, JavaScript

AI / ML

Claude Agent SDK, agentic evals & guardrails, RAG, LLM fine-tuning, LLaMA, HuggingFace Transformers, LangChain, PyTorch, TensorFlow

Data Engineering

Dataswarm, Presto, Spark, Hive, AWS, Docker, Kubernetes

Databases

PostgreSQL, Tao, MySQL, DynamoDB, MongoDB, Elasticsearch, Neo4j

Web & Visualization

React, React Native, Streamlit