Emmanuel A.

Key Skills

Software

No software listed

Top Subject Matter

E-Commerce - Product Categorization & Customer

Natural Language Processing (NLP)

Speech & Audio Processing

Top Data Types

Image

Video

Text

Top Task Types

Computer Programming/Coding

Data Collection

Object Detection

Question Answering

Text Generation

Fine-tuning

Freelancer Overview

Over the past three years, I have built a robust career in AI training data annotation, progressing from Junior AI Data Annotator to Senior Data Annotator across image, text, audio, video, and medical domains. At NeuralForge Data Solutions, I led high-stakes autonomous vehicle perception projects, annotating over 150,000 image frames with 2D/3D bounding boxes and instance segmentation masks while maintaining a 95% accuracy rate. I refined annotation guidelines for complex edge cases, such as partial occlusion and motion blur which reduced ambiguous labels from 12% to 3% and improved inter-annotator agreement from 0.72 to 0.91. I also mentored a team of eight junior annotators on DICOM medical imaging protocols and collaborated directly with ML engineers to refine label definitions, directly contributing to a model mAP improvement from 40% to 90%. Prior to this, at CogniLabel AI Labs, I specialized in NLP and speech data, processing 200,000+ text documents for NER and sentiment analysis and transcribing 5,000+ hours of multilingual audio (English, Pidgin, Hausa) for ASR models with 95% transcription accuracy. I developed Python automation scripts that cut pre-processing time by 60%, created company-wide style guides for low-resource African languages, and conducted weekly QA audits that boosted dataset F1-scores by 8 percentage points. Through freelance and academic projects including a medical pathology annotation initiative that achieved 94% pathologist-verified accuracy and an African language ASR corpus that improved baseline WER by 15%, I have consistently demonstrated the ability to deliver large-scale, high-quality training data that drives measurable model performance gains.

Labeling Experience

Education

Work History