Carren Kirwa

Key Skills

Software

CVAT

Scale AI

Remotasks

Appen

Labelbox

Top Subject Matter

STEM (math, physics, chemistry, statistics) and industrial manufacturing images

Stem Domain Expertise

data science

Top Data Types

Image

Text

Audio

Video

Document

Top Task Types

Bounding Box

Prompt + Response Writing (SFT)

Transcription

Freelancer Overview

Scientific Data Annotation Specialist — Handshake. Brings 9+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include CVAT, Outlier, and Scale AI. Education includes Doctor of Philosophy, University of Lincoln (2021) and Master of Science, University of Sunderland (2020). AI-training focus includes data types such as Image, Text, and Audio and labeling workflows including Bounding Box, Evaluation, and Rating.

Labeling Experience

Scientific Data Annotation Specialist — Handshake

CVATImageBounding Box

Reviewed and corrected auto-annotated frames from industrial and manufacturing image datasets using CVAT, focusing on process plant and refinery environments. Designed and validated STEM evaluation scenarios for mathematics, chemistry, thermodynamics, quantum mechanics, and statistics content. Authored and verified SFT prompt-and-response pairs in STEM and data science, and performed quality rating of GenAI outputs across multiple domains. • Applied bounding box annotation and classification labels using CVAT • Performed quality control and QA reviews of large-scale image datasets • Validated complex, multi-step mathematical and statistical reasoning in AI-generated solutions • Conducted red-team exercises for hallucination detection and error documentation

Reviewed and corrected auto-annotated frames from industrial and manufacturing image datasets using CVAT, focusing on process plant and refinery environments. Designed and validated STEM evaluation scenarios for mathematics, chemistry, thermodynamics, quantum mechanics, and statistics content. Authored and verified SFT prompt-and-response pairs in STEM and data science, and performed quality rating of GenAI outputs across multiple domains. • Applied bounding box annotation and classification labels using CVAT • Performed quality control and QA reviews of large-scale image datasets • Validated complex, multi-step mathematical and statistical reasoning in AI-generated solutions • Conducted red-team exercises for hallucination detection and error documentation

2026 - Present

STEM Domain Expert Evaluator — Outlier & TELUS Digital AI Community

Text

Assessed AI-generated text outputs for advanced STEM, data science, and ML tasks including equation solving, machine learning evaluation, and data science workflows. Performed side-by-side preference ranking and detailed rationale reporting for scientific accuracy and code correctness. Designed and validated computational statistics and programming problems, and authored multi-turn tutoring dialogues. • Evaluated GenAI outputs, including ranking and structured scoring • Validated solutions using Python-based workflows for reproducibility • Designed original analytical content and scenarios • Identified error patterns and provided structured feedback for model training

Assessed AI-generated text outputs for advanced STEM, data science, and ML tasks including equation solving, machine learning evaluation, and data science workflows. Performed side-by-side preference ranking and detailed rationale reporting for scientific accuracy and code correctness. Designed and validated computational statistics and programming problems, and authored multi-turn tutoring dialogues. • Evaluated GenAI outputs, including ranking and structured scoring • Validated solutions using Python-based workflows for reproducibility • Designed original analytical content and scenarios • Identified error patterns and provided structured feedback for model training

2023 - 2025

Annotation QA Lead & Guideline Author — Scale AI & Appen

Scale AIText

Led QA and guideline authorship for annotation projects across STEM, data science, ML, computer science, biology, and general domains. Designed gold-standard, honeypot, and calibration tasks while monitoring inter-annotator agreement and retraining needs. Executed QA sweeps of annotation files and collaborated with project leads to define rubric-based evaluation. • Wrote guidelines, appendices, and decision trees for annotation clarity • Designed calibration and gold-standard tasks for scientific content • Executed quality assurance sweeps using Python scripts • Reported throughput and quality metrics to platform leads

Led QA and guideline authorship for annotation projects across STEM, data science, ML, computer science, biology, and general domains. Designed gold-standard, honeypot, and calibration tasks while monitoring inter-annotator agreement and retraining needs. Executed QA sweeps of annotation files and collaborated with project leads to define rubric-based evaluation. • Wrote guidelines, appendices, and decision trees for annotation clarity • Designed calibration and gold-standard tasks for scientific content • Executed quality assurance sweeps using Python scripts • Reported throughput and quality metrics to platform leads

2022 - 2024

Speech & Text Data Annotator — Remote (Contract)

AppenAudioTranscription

Transcribed audio recordings, performed speaker diarization with speaker IDs, and annotated timestamps for non-speech events. Labeled text and chat logs for sentiment, toxicity, named entities, and relationships to train NER models. Reviewed ASR model outputs against reference transcripts to categorise error types for targeted model improvements. • Applied intent, topic, and toxicity labels to text and chat logs • Annotated named entities and conducted relationship labeling for NER • Conducted transcription and speaker diarization for audio data • Categorised and flagged ASR errors for QA and model improvement

Transcribed audio recordings, performed speaker diarization with speaker IDs, and annotated timestamps for non-speech events. Labeled text and chat logs for sentiment, toxicity, named entities, and relationships to train NER models. Reviewed ASR model outputs against reference transcripts to categorise error types for targeted model improvements. • Applied intent, topic, and toxicity labels to text and chat logs • Annotated named entities and conducted relationship labeling for NER • Conducted transcription and speaker diarization for audio data • Categorised and flagged ASR errors for QA and model improvement

2020 - 2023

AI Trainer & Data Labeler — Remotasks & DataAnnotation.tech

RemotasksTextPrompt Response Writing SFT

Authored SFT prompt-and-response pairs in Python, data science, statistical analysis, ML, SQL, math, and scientific reasoning domains. Performed image-based QA on scientific figures and label verification, and designed computational data science problems for validation. Conducted red-team and adversarial prompt tasks across STEM and data science scenarios. • Wrote multi-turn dialogue prompts for STEM tutoring contexts • Validated and flagged unanswerable or ambiguous examples • Ensured deterministic and reproducible problem solutions • Documented adversarial output reproduction for modeller review

Authored SFT prompt-and-response pairs in Python, data science, statistical analysis, ML, SQL, math, and scientific reasoning domains. Performed image-based QA on scientific figures and label verification, and designed computational data science problems for validation. Conducted red-team and adversarial prompt tasks across STEM and data science scenarios. • Wrote multi-turn dialogue prompts for STEM tutoring contexts • Validated and flagged unanswerable or ambiguous examples • Ensured deterministic and reproducible problem solutions • Documented adversarial output reproduction for modeller review

2021 - 2022

Show more

Education

Work History