Suzette Annan

Key Skills

Software

CrowdSource

Data Annotation Tech

OneForma

Scale AI

Other

Top Subject Matter

LLM evaluation in English.

Audio transcription, proofreading, and quality assurance.

RLHF, fine-tuning, and prompt writing.

Top Data Types

Audio

Image

Text

Top Task Types

Evaluation Rating

Fine Tuning

Prompt Response Writing SFT

RLHF

Freelancer Overview

I've worked for several LLM training platforms including Neevo, DataAnnotation.tech, and Outlier. I'm highly skilled in RLHF, particularly prompt writing and model response evaluation. I enjoy working on projects that utilise my critical thinking skills where I'm able to highlight nuances in language, bias, and sensitive topics. I also have experience in audio transcription.

Labeling Experience

Instruction Following

Data Annotation TechTextRLHFFine Tuning

This project involves evaluating the models' ability to follow the user's instructions and parameters in the prompt.

This project involves evaluating the models' ability to follow the user's instructions and parameters in the prompt.

2024

Prompt Writing

OtherTextEvaluation RatingPrompt Response Writing SFT

This project is on Scale AI's platform 'Outlier'. Tasks include creating challenging, hyper-specific prompts that cause the model to fail in at least one evaluation dimension. Some of the dimensions are truthfulness, instruction following, verbosity, and safety. The objective is to highlight and evaluate the weaknesses of the most advanced LLMs.

This project is on Scale AI's platform 'Outlier'. Tasks include creating challenging, hyper-specific prompts that cause the model to fail in at least one evaluation dimension. Some of the dimensions are truthfulness, instruction following, verbosity, and safety. The objective is to highlight and evaluate the weaknesses of the most advanced LLMs.

2024

Fact-checking

Data Annotation TechTextFine TuningEvaluation Rating

I work on several projects with the objective of ensuring that the models only generate content that is helpful, harmless and honest. This requires scrupulous research, fact-checking of the models' claims, and flagging model hallucinations.

I work on several projects with the objective of ensuring that the models only generate content that is helpful, harmless and honest. This requires scrupulous research, fact-checking of the models' claims, and flagging model hallucinations.

2024

Sensitive Topics

Data Annotation TechTextRLHFFine Tuning

I was required to compare and contrast the manner in which two models approached the same topic from the perspective of someone with a protected characteristic such as race, gender, sexuality, ethnicity, and religion etc. The project's objective was to highlight nuanced differences between model responses that indicated bias.

I was required to compare and contrast the manner in which two models approached the same topic from the perspective of someone with a protected characteristic such as race, gender, sexuality, ethnicity, and religion etc. The project's objective was to highlight nuanced differences between model responses that indicated bias.

2024

Audio Transcription

OtherAudioText Generation

I've worked on several transcription projects on the platform Neevo. Tasks include listening to audio clips in English and creating a written transcript. Other projects include listening to snippets of recorded conversations between non-native English speakers. My role was to evaluate the audibility and quality of the spoken English in the clips. I have experience in both verbatim and non-verbatim transcription.

I've worked on several transcription projects on the platform Neevo. Tasks include listening to audio clips in English and creating a written transcript. Other projects include listening to snippets of recorded conversations between non-native English speakers. My role was to evaluate the audibility and quality of the spoken English in the clips. I have experience in both verbatim and non-verbatim transcription.

2022

Show more

Education

Work History