Sara Ghorayeb

AI engineer, London

SaraGhorayeb

I evaluate AI agents for a living, I train for triathlon, and my degree is in molecular bioengineering. This site keeps one theme for each: evals, triathlon and healthcare.

I lead agent evaluation at GSK, where I built the evaluation infrastructure behind every release and pull request of a biomedical research agent that 3,000 scientists use. Evals holds six projects with their results. Triathlon and healthcare hold none yet, and their pages say so.

Role
AI Engineer, Responsible AI, GSK
Location
London, UK
Languages
English, French, Arabic
Email
saraghorayeb2@gmail.com
GitHub
github.com/saraghorayeb29

Three themes

filled light: the theme has projects / hollow light: no project yet

Theme 16 projects

Evals

I build evaluation infrastructure for AI agents, and I measure whether their judges agree with people.

One result of six

Cohen's kappa of an LLM judge against the human majority label, with its 95% interval. 121 MT-Bench items, the same items for both judges. The intervals overlap, so this sample does not establish a difference.

The live judge console and the six projects
Theme 2No project yet

Triathlon

I train for triathlon.

Three empty lanes, one per discipline of the sport. Nothing is recorded on them. The panel shows no number because I have published none.

The triathlon page
Theme 3No project yet

Healthcare

I studied molecular bioengineering, and the agent I evaluate at work is a biomedical research agent.

An empty frame. No project is in it. The rows below are my education and one manuscript, and they are not projects of this site.

  • 00Projects: none yet. I have not scoped the first one.
  • EdMEng Molecular Bioengineering, First Class Honours, Imperial College London
  • MsChest X-ray counterfactuals, a manuscript I co-authored, under review at ICLR 2027
The healthcare page

Experience and education

2025 to present

AI Engineer, Responsible AIGSK, London, UK

2023 to 2025

Future Leaders Programme: Data Scientist and Cyber Security Analyst rotationsGSK, London, UK

Jul to Sep 2022

Software Engineer InternBios Health, Cambridge, UK

2019 to 2023

MEng (Master's) Molecular Bioengineering, First Class HonoursImperial College London, London, UK

Research

Two manuscripts. Neither is published yet. One belongs to evals and one to medical imaging.

First authorIn preparation

Deep Evaluator: Turning Agent Evaluation into Development Evidence

My first-author research is on agentic judges, judge calibration and causal failure analysis. The experiment code is described on the Deep Evaluator page.

Co-authorUnder review, ICLR 2027

Generating Chest X-Ray Counterfactuals by Specialising Foundation Image Models

It matches a 766M-parameter model trained from scratch by fine-tuning 24M LoRA parameters.

03  Contact

Emailsaraghorayeb2@gmail.com
GitHub
github.com/saraghorayeb29
Location
London, UK
Languages
English, French, Arabic