Teresia Mumbua

Key Skills

Software

Scale AI

Labelbox

Appen

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Text

Image

Audio

Video

Top Label Types

Classification

RLHF

Evaluation Rating

Prompt Response Writing SFT

Bounding Box

Segmentation

Object Detection

Emotion Recognition

Audio Recording

Transcription

Action Recognition

Tracking

Freelancer Overview

I am a detail-oriented AI data reviewer with over six years of hands-on experience in structured annotation, quality assurance, and human-in-the-loop AI training pipelines, primarily focused on NLP tasks such as text classification, summarization, and complex reasoning. My work emphasizes high-precision workflows, where I consistently maintain 99%+ accuracy across large-scale datasets and excel at identifying edge cases, ambiguity, and rubric misalignment to improve downstream model performance. I am highly skilled in using platforms like Labelbox, Scale AI, and various proprietary annotation tools, as well as leveraging Google Sheets, Excel, and basic SQL for quality tracking and data validation. My experience includes leading calibration projects to improve scoring consistency, developing standardized feedback frameworks, and optimizing labeling workflows to reduce correction cycles. I thrive in independent, remote environments and am committed to delivering guideline-compliant, structured feedback that enhances AI output reliability and evaluation consistency.

Labeling Experience

Computer Vision Image Annotation for Object Detection

LabelboxImageBounding BoxSegmentation

Annotated large image datasets for machine learning models used in computer vision applications. Tasks included drawing bounding boxes around objects, labeling object categories, and performing segmentation to identify precise object boundaries. Ensured accurate annotation of objects across varying lighting conditions, occlusions, and complex backgrounds. Followed strict labeling guidelines to maintain dataset consistency and improve model detection accuracy. Contributed to the preparation of high-quality datasets used to train object detection algorithms capable of recognizing multiple objects in real-world environments.

Annotated large image datasets for machine learning models used in computer vision applications. Tasks included drawing bounding boxes around objects, labeling object categories, and performing segmentation to identify precise object boundaries. Ensured accurate annotation of objects across varying lighting conditions, occlusions, and complex backgrounds. Followed strict labeling guidelines to maintain dataset consistency and improve model detection accuracy. Contributed to the preparation of high-quality datasets used to train object detection algorithms capable of recognizing multiple objects in real-world environments.

2021 - 2025

AI Training Data Annotation for Large Language Models

Scale AITextClassificationRLHF

Worked on large-scale data labeling projects designed to train and evaluate advanced AI language models. Responsibilities included reviewing prompt-response pairs, rating AI-generated outputs for accuracy, relevance, and safety, and identifying edge cases or inconsistencies in model behavior. Performed tasks such as text classification, response evaluation, prompt generation, and reinforcement learning from human feedback (RLHF) to improve model performance. Maintained strict adherence to detailed annotation guidelines to ensure high-quality training data. The project involved evaluating thousands of data samples while maintaining consistency, precision, and quality assurance standards required for machine learning training pipelines. Special focus was placed on detecting ambiguity, bias, and factual inaccuracies in AI-generated responses.

Worked on large-scale data labeling projects designed to train and evaluate advanced AI language models. Responsibilities included reviewing prompt-response pairs, rating AI-generated outputs for accuracy, relevance, and safety, and identifying edge cases or inconsistencies in model behavior. Performed tasks such as text classification, response evaluation, prompt generation, and reinforcement learning from human feedback (RLHF) to improve model performance. Maintained strict adherence to detailed annotation guidelines to ensure high-quality training data. The project involved evaluating thousands of data samples while maintaining consistency, precision, and quality assurance standards required for machine learning training pipelines. Special focus was placed on detecting ambiguity, bias, and factual inaccuracies in AI-generated responses.

2021 - 2025

Speech Data Annotation and Transcription for AI Voice Recognition Systems

AppenAudioEmotion RecognitionAudio Recording

Contributed to speech data annotation projects used to train automatic speech recognition (ASR) and conversational AI systems. Tasks included transcribing spoken audio into accurate text, labeling emotional tone in speech recordings, and reviewing audio clips for clarity, pronunciation, and speaker intent. Ensured high transcription accuracy by carefully identifying accents, background noise, and overlapping speech. Followed strict annotation guidelines to maintain dataset consistency and improve AI models’ ability to understand natural human speech across different contexts. Worked with large batches of audio recordings, verifying transcription quality and correcting inconsistencies to support the development of reliable voice-enabled AI technologies.

Contributed to speech data annotation projects used to train automatic speech recognition (ASR) and conversational AI systems. Tasks included transcribing spoken audio into accurate text, labeling emotional tone in speech recordings, and reviewing audio clips for clarity, pronunciation, and speaker intent. Ensured high transcription accuracy by carefully identifying accents, background noise, and overlapping speech. Followed strict annotation guidelines to maintain dataset consistency and improve AI models’ ability to understand natural human speech across different contexts. Worked with large batches of audio recordings, verifying transcription quality and correcting inconsistencies to support the development of reliable voice-enabled AI technologies.

2023 - 2024

Video Annotation and Action Recognition for AI Vision Systems

CVATVideoObject DetectionAction Recognition

Participated in video data annotation projects used to train and evaluate computer vision models. The work involved reviewing video sequences and labeling objects frame-by-frame to help AI systems recognize activities and track moving objects over time. Tasks included identifying and tagging human actions, detecting objects in motion, and performing object tracking across consecutive frames. Ensured consistent annotations across long video sequences by following strict labeling guidelines and quality control standards. The annotated datasets supported the development of AI systems capable of understanding dynamic visual environments and recognizing actions in real-world scenarios.

Participated in video data annotation projects used to train and evaluate computer vision models. The work involved reviewing video sequences and labeling objects frame-by-frame to help AI systems recognize activities and track moving objects over time. Tasks included identifying and tagging human actions, detecting objects in motion, and performing object tracking across consecutive frames. Ensured consistent annotations across long video sequences by following strict labeling guidelines and quality control standards. The annotated datasets supported the development of AI systems capable of understanding dynamic visual environments and recognizing actions in real-world scenarios.

2022 - 2023

Education

Work History