Elisha Jackson

Reviewed generated images for instruction adherence and visual integrity, maintaining a 98% accuracy rate across large annotation batches for computer vision training. Used image annotation/markup tooling to complete 50+ high-complexity image evaluations per day.

Performed dense and coarse video/text labeling, achieving a 95% first-pass approval rating from senior quality reviewers. Reviewed AI-generated descriptions for grammatical accuracy and visual alignment at ~40 video-text pairs per shift and updated process execution across 3 major guideline changes without production speed loss.

Evaluated high-volume image datasets for technical quality factors, identifying artifacts and noise across batches of 200+ images per shift. Applied standardized scoring, filtered low-quality images across 15+ defect types, and contributed to a 5% reduction in downstream training quality issues.

Classified multimodal content for safety and policy compliance by auditing 1,000+ units to identify misinformation and abusive behavior. Evaluated model outputs for factual reliability and worked to reduce hallucinations in assigned datasets with evidence-based feedback, ranking in the top 15% of annotators.

Reviewed and compared 800+ AI-generated responses for accuracy, clarity, reasoning quality, and instruction adherence on complex reasoning tasks. Authored and tested 150+ custom prompts to surface edge cases, documenting 20+ recurring issues that contributed to a 10% increase in internal QA alignment.

Designed multi-turn conversations that simulated realistic user interactions with virtual assistants across email, calendar, maps, and cloud storage workflows. Evaluated assistant outputs for accuracy, consistency, formatting compliance, and natural communication while following detailed project guidelines.

Earned a B.Sc. in Computer Science from the University of Uyo, graduating in 2024.