Looking for a job

Philomena Mumbua

@philomenamumbua

Message

AI Data Annotator at Sama who reviewed text, image, audio, video, and LiDAR data for AI projects.

Kenya

Message

At Sama, I annotated and reviewed text, image, audio, video, and LiDAR data for AI and machine-learning projects. I handled 2D/3D labeling, segmentation, object tracking, and language tasks, following project guidelines and escalating unclear examples.

At Aventure, I supported LLM optimization through prompt experimentation, model-output evaluation, and preference ranking, with a focus on Kiswahili and English. At Code for Africa (iLAB), I collected, cleaned, annotated, and quality-checked multilingual datasets, using Python workflows to process 120,000+ records while maintaining approximately 98% quality.

Experience

Work history, roles, and key accomplishments

AI Data Trainer / Multimodal Annotator

Sama

Jan 2022 - Dec 2025 (3 years 11 months)

Annotated and reviewed text, image, audio, video, and LiDAR data for AI and machine-learning projects. Performed 2D/3D labeling, segmentation, object tracking, transcription, NER, sentiment, intent, and dialogue annotation.

Data Annotation Image Annotation Video Annotation LiDAR Segmentation Object Tracking Transcription NER Sentiment Analysis Intent Labeling Dialogue Labeling Quality Assurance

AI Consultant / Language & Data Specialist

Aventure

Jan 2024 - Dec 2024 (11 months)

Supported LLM optimization through prompt experimentation, model-output evaluation, and preference ranking. Reviewed multilingual language data for quality and cultural appropriateness, documenting errors and providing recommendations.

LLM Evaluation Prompt Engineering Preference Ranking Data Quality Kiswahili English Error Analysis

Research & Data Annotation Fellow

Code for Africa (iLAB)

Jan 2025 - Present (1 year 9 months)

Collected, cleaned, annotated, and quality-checked multilingual media and research datasets across African markets. Applied structured taxonomies and used Python workflows to handle large datasets while maintaining high quality.