abigailhaddad/compliancebenchmark

Contact

• Email
• LinkedIn
• Substack

Bio

Hi, I'm Abigail Haddad. I'm a data scientist working on LLM evals. I previously worked on research and data science for the Department of Defense, including at the RAND Corporation and as a Department of the Army civilian.

My areas of interest include:

• Methods for evaluating/classification of text data, including via LLMs
• Government data workforce/talent management issues
• Code-first data science and processes for better transparency and replication
• DC school data

Get in touch if you're interested in getting involved with/sponsoring Data Community DC or having me speak at your event.

Upcoming Talks

• "Automating Tests for your RAG Chatbot or Other Generative Tool" - New York R Conference, May 16-17
  • Conference Site

Recent Talks

A Framework for Testing LLMs for Outputting Hazardous Information - Generative AI DC

• Video

Automated Evaluations for Your RAG Chatbot or Other Generative Tool - AI In Production

• Video

What Job Is This, Anyway?: Using LLMs to Classify USAJobs Data Scientist Listings - RGOV Conference 2023

• Video
• Slides and Notes
• Code

Writing

• Can Mistral Large Do That? Using an LLM Grader to Quickly Evaluate New Models
• Mapping Testing to Threats: For pre-release testing, the threat model should inform the methods
• Inferring numerators and denominators from decimals using SymPy to re-identify anonymized data
  • Code
• Improving Your Junior Data Scientist Resume
• Improving Your Jupyter Notebook GitHub Portfolio