Maximus Makokha

Key Skills

Software

Appen

Labelbox

CVAT

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Text

Image

Video

Top Task Types

Translation/Localization

Evaluation/Rating

Polygon

Point/Key Point

Segmentation

Text Generation

Text Summarization

Freelancer Overview

I am a detail-oriented data annotator and linguistic evaluator with hands-on experience in both audio transcription and image annotation for AI and machine learning projects. My background includes transcribing and reviewing English and Swahili audio, annotating agricultural crop images using Labelbox, and evaluating translation consistency for NLP tasks. I excel at ensuring data accuracy, following strict project guidelines, and using tools like Microsoft Office, Google Workspace, and basic Python for text automation. I am passionate about contributing high-quality, well-labeled datasets that support the development of reliable AI models, and I am comfortable working independently or as part of a remote team to meet demanding deadlines.

Labeling Experience

Egocentric Annotation Program

Internal Proprietary ToolingVideoSegmentationText Generation

This project involves human egocentric video annotation focused on identifying and segmenting physical actions performed from a first-person (ego) perspective. The scope of the project includes reviewing videos captured by wearable or first-person cameras and dividing them into distinct action segments that represent meaningful task steps performed by the human operator. Tasks performed include accurately identifying action boundaries, labeling segments based on observed human activities, and ensuring each segment represents a single, coherent action within a broader task workflow. Annotations are created following detailed project guidelines to maintain consistency across videos and environments. Quality measures include strict adherence to segmentation rules, careful review of action transitions, avoidance of overlapping or mixed segments, and self-validation of annotations prior to submission. The project supports the development of computer vision and embodied AI systems.

This project involves human egocentric video annotation focused on identifying and segmenting physical actions performed from a first-person (ego) perspective. The scope of the project includes reviewing videos captured by wearable or first-person cameras and dividing them into distinct action segments that represent meaningful task steps performed by the human operator. Tasks performed include accurately identifying action boundaries, labeling segments based on observed human activities, and ensuring each segment represents a single, coherent action within a broader task workflow. Annotations are created following detailed project guidelines to maintain consistency across videos and environments. Quality measures include strict adherence to segmentation rules, careful review of action transitions, avoidance of overlapping or mixed segments, and self-validation of annotations prior to submission. The project supports the development of computer vision and embodied AI systems.

2025

Karimata English Swahili translation evaluation

AppenTextTranslation Localization

This project involves evaluating large language model (LLM) translation outputs between English and Swahili using the Appen platform. The scope of work includes reviewing AI-generated translations for accuracy, fluency, grammatical correctness, contextual meaning, and cultural appropriateness. Tasks performed include comparing source and target texts, identifying mistranslations, omissions, or unnatural phrasing, and assigning quality ratings on a 1–5 scale, where 5 represents a near-perfect translation and 1 represents a poor or incorrect translation. Evaluations are conducted according to predefined guidelines to ensure consistent and reliable feedback for model improvement. Quality measures include strict adherence to evaluation rubrics, careful cross-checking of meaning between languages, consistency in scoring, and self-review to minimize subjectivity and rating drift.

This project involves evaluating large language model (LLM) translation outputs between English and Swahili using the Appen platform. The scope of work includes reviewing AI-generated translations for accuracy, fluency, grammatical correctness, contextual meaning, and cultural appropriateness. Tasks performed include comparing source and target texts, identifying mistranslations, omissions, or unnatural phrasing, and assigning quality ratings on a 1–5 scale, where 5 represents a near-perfect translation and 1 represents a poor or incorrect translation. Evaluations are conducted according to predefined guidelines to ensure consistent and reliable feedback for model improvement. Quality measures include strict adherence to evaluation rubrics, careful cross-checking of meaning between languages, consistency in scoring, and self-review to minimize subjectivity and rating drift.

2025

Agricultural crop annotation

CVATImagePolygonPoint Key Point

This project focuses on image annotation for agricultural crop and weed identification using CVAT. The scope of work involves labeling field and close-range agricultural images by creating precise polygon annotations to outline crops and weeds, as well as point/key point annotations to mark specific plant features and reference locations. Tasks performed include differentiating crops from weeds, accurately tracing object boundaries, assigning correct class labels based on provided taxonomies, and handling edge cases such as overlapping vegetation or partially visible plants. Quality measures include strict adherence to annotation guidelines, manual self-review of all annotations, consistency checks across image sets, and correction of feedback to maintain high labeling accuracy for machine learning model training.

This project focuses on image annotation for agricultural crop and weed identification using CVAT. The scope of work involves labeling field and close-range agricultural images by creating precise polygon annotations to outline crops and weeds, as well as point/key point annotations to mark specific plant features and reference locations. Tasks performed include differentiating crops from weeds, accurately tracing object boundaries, assigning correct class labels based on provided taxonomies, and handling edge cases such as overlapping vegetation or partially visible plants. Quality measures include strict adherence to annotation guidelines, manual self-review of all annotations, consistency checks across image sets, and correction of feedback to maintain high labeling accuracy for machine learning model training.

2025

Agricultural crop annotation

LabelboxImagePolygonPoint Key Point

This project involves agricultural image annotation for crop labeling and identification using Labelbox. The scope of work includes annotating high-resolution field and aerial images by creating polygon annotations to accurately outline crop boundaries and point/key point labels to mark specific crop features and reference points. Tasks performed include precise polygon drawing around crop areas, key point placement for crop identification, and consistent class labeling according to project guidelines. Annotations are completed following strict quality standards, including visual verification, boundary accuracy checks, and adherence to labeling instructions to ensure high-quality training data for computer vision models used in agricultural analysis and monitoring.

This project involves agricultural image annotation for crop labeling and identification using Labelbox. The scope of work includes annotating high-resolution field and aerial images by creating polygon annotations to accurately outline crop boundaries and point/key point labels to mark specific crop features and reference points. Tasks performed include precise polygon drawing around crop areas, key point placement for crop identification, and consistent class labeling according to project guidelines. Annotations are completed following strict quality standards, including visual verification, boundary accuracy checks, and adherence to labeling instructions to ensure high-quality training data for computer vision models used in agricultural analysis and monitoring.

2025

Karimata

AppenTextTranslation LocalizationEvaluation Rating

The Karimata Evaluation Project focuses on assessing and improving the quality, accuracy, and safety of AI-generated outputs. The project involves systematically reviewing model responses against defined evaluation criteria, identifying errors, inconsistencies, bias, or policy violations, and providing structured feedback to support model refinement. Key responsibilities include evaluating AI responses for correctness, relevance, clarity, and alignment with guidelines; classifying outputs using established taxonomies; and documenting findings in a clear and reproducible manner. The project emphasizes attention to detail, consistency, and adherence to evaluation rubrics to ensure high-quality training data for AI model improvement.

The Karimata Evaluation Project focuses on assessing and improving the quality, accuracy, and safety of AI-generated outputs. The project involves systematically reviewing model responses against defined evaluation criteria, identifying errors, inconsistencies, bias, or policy violations, and providing structured feedback to support model refinement. Key responsibilities include evaluating AI responses for correctness, relevance, clarity, and alignment with guidelines; classifying outputs using established taxonomies; and documenting findings in a clear and reproducible manner. The project emphasizes attention to detail, consistency, and adherence to evaluation rubrics to ensure high-quality training data for AI model improvement.

2025

Education

Work History